Search

Over the past decade, the advancements in technology and the growing amount of identified genetic variants have led to a high number of important discoveries in the field of precision medicine concerning human biology and pathophysiology. However, it became evident that genomics alone could not properly explain the onset and regulation of the specific molecular mechanisms of certain phenotypes. Studying omics helped complement this gap in genetic research, providing detailed information on the quantification of molecules that are involved in structural and functional processes in the organism. Specifically, protein production, levels, and regulation are dynamic and change during the course of one's lifetime. This information has proven fundamental to understanding how certain proteins affect complex phenotypes such as neurological and psychiatric disorders. In this thesis, I describe the three groups of analyses I conducted over the course of my doctoral programme on different sets of blood plasma proteins and over a broad range of neurological, psychiatric, cardiovascular, and electrophysiology phenotypes. The underlying mechanisms that trigger the onset of psychiatric and neurological conditions are often not limited to the nervous system, but rather stem from multi-system molecular triggers. The first part of the work I carried out aims at investigating the frequent co-occurrence and comorbidity of neurological and cardiovascular phenotypes by conducting a genome-wide association (GWA) meta-analysis of 183 neurology-related blood proteins on data from over 12000 individuals. The second part concerns the bivariate and multivariate analyses conducted on 276 cardiology and inflammatory proteins, while the third illustrates the contribution to consortia focussed on heart rate and electrophysiology. Results from the second and third parts of the work provided information that played an important role in understanding a part of the genetic mechanisms of the complex traits of interest. Overall, the results presented in this thesis strongly support the notion that proteomics is an important tool to be used to study complex traits and drug discovery and development should focus on targeting protein synthesis and regulation. Furthermore, the results also support the notion that complex diseases involve more than one biological system, and in order to gain a better understanding of human pathology, it is fundamental to study the causes and effects across the entire organism.

Precursor B-cell acute lymphoblastic leukaemia (pre-B ALL) is a haematological malignancy characterised by immature proliferation of B-cell lineage blasts. Current standard of care treatment for pre-B ALL involves multi-drug regimens of chemotherapy agents which despite high curative rates are often poorly tolerated by patients, inducing off-target side effects. To address this unmet clinical need, this project has identified for the first time a novel anticancer agent comprised of three phospholipase A2 (PLA2) enzymes. The compound, denominated PA-PLA2, was purified from Pseudechis australis snake venom using Reverse-Phase High Performance Liquid Chromatography, and validated by matrix assisted laser desorption ionisation-time of flight mass spectrometry and sequencing. Using a combination of cell-based functional assays and flow cytometry, PA-PLA2 was identified to induce caspase-independent necrotic cell death in two pre-B ALL cell lines representative of paediatric (Reh) and adult (SD-1) disease. Employing RNA-sequencing, protein- and phospho-proteomics analysis with time course Western blotting, elucidated PA-PLA2's mechanism of action to be RIPK1 and Akt dependent, inducing necroptosis in Reh cells and stunted cell cycle progression via Akt/p27Kip1 signalling in SD-1 cells. Importantly, PA-PLA2 treatment had minimal effects on the cell viability of a range of normal, peripheral blood and bone marrow-derived stem cells obtained from patient samples, with assays in donor citrated whole blood samples demonstrating PA-PLA2 to be non-haemolytic. Finally, by using a custom panel of 48 compounds, PA-PLA2 was identified to act synergistically in combination with the pre-B ALL induction phase chemotherapy agents, Cytarabine and Idarubicin. Collectively, the findings from this project support the future use of snake venom toxins in pre-B ALL treatment. The identification, validation and characterisation of PA-PLA2 represents a platform for future investigations using animal venoms as natural sources for therapeutic agents, with further clinical investigation in more clinically relevant models needed to translate this thesis' findings.

Advances in profiling technologies have enabled systematic assessment of a plethora of molecules from genome via proteome to a variety of metabolic phenotypes, providing unique opportunities to better understand disease mechanisms and identify molecular signatures that can help to predict, screen for and diagnose different diseases using agnostic discovery approaches. The proteome is at the centre of biological information transfer and therefore provides a unique source of actionable discoveries. The increasing global burden of Type 2 diabetes (T2D) calls for improved strategies to identify individuals at high-risk. However, the aetiology of this heterogenous disease remains incompletely understood, and some subgroups are poorly predicted or overlooked. The main aim of this thesis was to identify plasma proteomic signatures for improved risk prediction and aetiological understanding of T2D and related metabolic traits. I first conducted a systematic characterisation of the major phenotypic and genetic determinants of plasma protein levels, which showed a range of modifiable risk factors as causal determinants that improved interpretation of disease biomarkers and candidate mediators. To assess the integrated whole-body response to high glucose ingestion, I systematically characterised protein changes during an oral glucose tolerance test (OGTT). Around 11% of the measured proteome changed after the OGTT, pointing to groups of proteins with a differential response in insulin resistant compared to insulin sensitive individuals, and that were associated with long term cardiometabolic consequences. To go beyond understanding of molecular mechanisms associated or involved in glucose homeostasis and explore the predictive potential of the plasma proteome, I applied a machine learning framework to ~5,000 proteins measured at fasting. I identified a 3 protein signature that improved discrimination over clinical risk factors of isolated impaired glucose tolerance (iIGT), a group at high risk for T2D and comorbidities that is missed by current screening guidelines. Further systematic assessment showed that a sparse signature of features across the proteome, metabolome and genome improved risk prediction for incident T2D, of which the major driver was the T2D-polygenic score. However, individuals with HbA1c below the threshold for prediabetes but at high polygenic risk had a substantially lower cumulative T2D incidence than people with prediabetes. To investigate the predictive potential of plasma proteomics more broadly and disease-specificity of predictive biomarkers, I developed sparse protein models for the prediction of 24 diverse incident diseases across a range of clinical specialties. As few as five proteins improved the predictive performance of patient derived risk factors for seven diseases, revealing the potential of high-throughput proteomics for novel biomarker discovery and improved risk prediction across a range of diseases. This thesis demonstrates the potential of broad-capture plasma proteomics for identification of novel aetiological pathways and predictive biomarkers to improve or refine risk prediction and screening strategies, specifically for subgroups that are poorly captured by current clinical risk factors.

Age-related macular degeneration (AMD) is a leading cause of blindness in the elderly worldwide. Drusen, extracellular deposits of debris accumulating in Bruch's membrane, are considered hallmarks of early AMD. Dysregulation of the complement cascade is believed to be a driver in the pathogenesis of AMD. However, why the macular region is particularly susceptible to AMD remains unclear. In this thesis, I investigate differences in molecular composition of the macular and peripheral Bruch's membrane with associated choroidal extracellular matrix (enriched Bruch's membrane) from eyes donated by older individuals without AMD. Furthermore, I analyse and compare the localisation of complement proteins in eyes from older donors with and without early AMD. In this study, using isobaric tags for relative and absolute quantitation (iTRAQ) mass spectrometry, a total of 1,722 proteins were identified in enriched Bruch's membrane from donors who did not have AMD. Quantitative analysis revealed that 14 proteins were significantly more abundant in the macular region than in the peripheral region, including complement proteins (C3, C5, C6 and C7). Conversely, seven proteins were significantly more abundant in the peripheral region than in the macula. Subsequent immunohistochemistry studies demonstrated that C5b-9, FHR-4 and FB localised mainly in the intercapillary septa and around the capillaries of the choriocapillaris in AMD donors, but these were barely seen within the Bruch's membrane structure itself in non-AMD donors. However, in eyes with early AMD, these complement components were observed as an intense band throughout Bruch's membrane and within drusen. Overall, this study shows regional differences in the proteins of enriched Bruch's membrane, with an increased concentration of C3 and terminal pathway proteins in the macula compared to the periphery; this mirrors the macular predilection for AMD. Additionally, the translocation of complement proteins from the intercapillary septa to within Bruch's membrane itself between non-AMD and AMD donor eyes indicates that the complement pathway is initially activated in the intercapillary septa and immediately around the capillaries of the choriocapillaris. As the disease progresses Bruch's membrane becomes more permeable to complement proteins allowing them to permeate through its layers and enter drusen.

The cellular landscape changes dramatically over the course of a 24 h day. The proteome responds directly to daily environmental cycles and is additionally regulated by the circadian clock. Historically, models of the circadian clock have been built from transcriptomic studies, but over the past few years it has become clear that protein abundance information cannot be assumed from transcriptomic data. To quantify the relative contribution of diurnal versus circadian regulation, we mapped proteome dynamics under light:dark cycles compared with constant light. Using Ostreococcus tauri, a prototypical eukaryotic cell, we achieved 85% coverage, which allowed an unprecedented insight into the identity of proteins that facilitate rhythmic cellular functions. The overlap between diurnally- and circadian-regulated proteins was modest and these proteins exhibited different phases of oscillation between the two conditions. Transcript oscillations were generally poorly predictive of protein oscillations, in which a far lower relative amplitude was observed. We observed coordination between the rhythmic regulation of organelle-encoded proteins with the nuclear-encoded proteins that are targeted to organelles. Rhythmic transmembrane proteins showed a different phase distribution compared with rhythmic soluble proteins, indicating the existence of a circadian regulatory process specific to the biogenesis and/or degradation of membrane proteins. Our observations argue that the cellular spatiotemporal proteome is shaped by a complex interaction between intrinsic and extrinsic regulatory factors through rhythmic regulation at the transcriptional as well as post-transcriptional, translational, and post-translational levels. The most prominent extrinsic factor that might affect the proteome is our planet's 24h light/dark cycle and its effect on the generation of reactive oxygen species (ROS). Redox homeostasis and the circadian clock regulate one another to negate the potential effects of rhythmic light and attain homeostasis. Selenoproteins are an important class of redox-related enzymes that have a selenocysteine residue in the active site. We developed a functional understanding of how environmental and endogenous circadian rhythms integrate to shape the selenoproteome in a model eukaryotic cell. We mined quantitative proteomic data for the 24 selenoproteins of Ostreococcus across time series, again under environmentally rhythmic entrained conditions of light/dark (LD) cycles, compared to constant circadian conditions of constant light (LL). We found an overrepresentation of selenoproteins among rhythmic proteins under LL, but an underrepresentation under LD conditions. Rhythmic selenoproteins under LL that reach peak abundance later in the day showed a greater relative amplitude of oscillations than those that peak early in the day. Under LD, amplitude did not correlate with peak phase; however, we identified high-amplitude selenium uptake rhythms under LD but not LL conditions. Selenium deprivation induced strong qualitative defects in clock gene expression under LD but not LL conditions. Overall, the clear conclusion is that the circadian and environmental cycles exert differential effects on the selenoproteome, and that the combination of the two enables homeostasis. Selenoproteins may therefore play an important role in the cellular response to reactive oxygen species that form as a consequence of the transitions between light and dark. In both the overall proteome and the selenoproteome, the overlap between diurnally-and circadian-regulated proteins was small, and transcript rhythmicity was poorly predictive of protein abundance rhythms. We therefore sought to obtain a circadian RNA-binding proteome to investigate temporal post-transcriptional regulation. We developed a protocol to use UV at 254 nm to covalently crosslink proteins in direct contact with RNA, extract them and then identify the proteins. Alongside the anticipated ribosomal and known RNA-binding proteins, we identified key metabolic enzymes that rhythmically bind RNA under constant circadian conditions. Post-transcriptional regulation by (or of) moonlighting metabolic enzymes must therefore be considered as a potential link reciprocally connecting metabolism to the circadian clock. To ultimately reveal the integration of endogenous and environmental rhythms on the rhythmic flow of information from DNA to rhythmic function, genetic tools are indispensable. While the model cells of Ostreococcus have been instrumental in obtaining detailed knowledge of rhythmicity of the cellular landscape at the -omics level, genetic tools to take the next steps are underdeveloped for this organism. We have made important headway in the set-up of the CRISPR/Cas9 genome editing system in Ostreococcus, to make precise mutations in target gene sequences. We designed two methods for the positive selection of edited cells, and optimised the parameters for the reliable transformation of Cas9 ribonucleoprotein complexes into the algal cells. We designed a diagnostic PCR method for the identification of edited algal DNA, which showed that the genome editing of Ostreococcus can certainly be successful. This thesis provides a step towards an integrated -omics overview of circadian regulation of the eukaryotic cell. Additional temporally resolved datasets, such as for the phosphoproteome and metabolome, taking advantage of the minimal cellular complexity of Ostreococcus, should be collected to fully comprehend the complex interactions between the levels of physiology that together shape cellular rhythmicity.

Proteomics has reached a turning point where the datasets are now approaching population-scale, this was achieved via faster and more sensitive instruments, improved sample handling and robust software tools. The population-scale era creates the opportunity to further study the molecular differences across cell lines, cell types or tissues as well as phenotypes of health and disease in human patients. In this thesis we utilize different methodologies to perform large-scale proteomic analyses, including a data dependent acquisition (DDA) based dataset using tandem mass tags (TMT) and a label-free data independent acquisition (DIA) one. The two studies looked at different systems, including induced pluripotent stem cells (hiPSCs) as well as peripheral blood neutrophils derived from both healthy and diseased patients and provide new biological insights including new disease specific biomarkers as well as novel drug targets. On a technical aspect, this work revealed novel considerations and limitations of using TMT for large-scale proteomic experiments, providing a framework to improve future studies and minimise the potential issues involved. Furthermore, the TMT-based hiPSC dataset revealed novel protein level effects caused by the erosion of X chromosome inactivation (XCI) in healthy female hiPSCs. The data show the erosion of XCI increases abundance of not just X chromosome proteins, but over 2,000 proteins derived from all other chromosomes, thereby significantly increasing the protein content in these eroded female lines compared to male and non-eroded female lines. The DIA-based large-scale proteomic characterisation of neutrophils derived from control and COVID19 patients revealed the effects of SARS-CoV2 infection in human neutrophils at the early infection and recovery phase. It highlighted a core signature present in the early infection timepoints in COVID19 patients, as well as some transient and some persistent changes on the neutrophil proteomes caused by COVID19. The study also highlighted the potential for patient stratification and precision medicine, as it detected important proteomic changes that were only present in patients with critically severe COVID19, and empowered treatment options that could help improve clinical trajectories for these patients. Furthermore, the data provided molecular insights into the delayed recovery state seen in a subset of COVID19 patients, as this patient group displayed a dysfunctional neutrophil phenotype reminiscent of what is seen in chronic diseases like COPD. Finally, this thesis explores the value of web applications (web app) to visualise and explore the proteomic data. Proteomic datasets are growing in size and complexity, thus finding ways to share and visualise this complex data in an intuitive format for non-specialist users becomes an important goal. As a solution we built two web apps, the Encyclopedia of Proteome Dynamics (EPD) and the Immunological Proteome Resource (ImmPRes), to ensure open access to the to the raw and processed proteomic data, as well as easy visualisation/exploration via graphical interfaces. ImmPRes provides access to the biggest collection of leukocyte proteomic data, integrating data from innate and adaptive cells as well as multiple datasets characterising different T cell populations along with multiple signalling pathways that modulate their functions, all contained within a simple web application that is easily accessible for all immunologists.

Type-2 Diabetes Mellitus (T2DM) is a major global concern that affects millions of people worldwide. An estimate of 35% of individuals with T2DM do not derive benefit from metformin therapy and an average of 40% of diabetic individuals do not derive benefit from GLP-1 therapy. Prolonged elevated HbA1c levels in patients contributes to the development of detrimental secondary complications. Comorbidities are common in individuals affected with T2DM with 88.6% of individuals in a survey held in 2004 to have one comorbid condition while 15% of individuals to have four or more comorbid conditions. The objective of this thesis is: (i) To identify and understand biomarkers associated with treatment response and comorbidities in T2DM using a clinical, genomic and proteomic approach; (ii) To identify and understand subgroups that respond to specific treatment interventions in the case of metformin monotherapy and GLP-1 therapy and in comorbidities of T2DM; (iii) To develop Machine Learning based models that can effectively stratify treatment groups, non-response groups and comorbid groups before diagnosis using a genomic approach and after diagnosis using clinical and proteomic approach. This thesis reports analysis(s) from a T2DM dataset from the Diastrat cohort recruited at the Northern Ireland Centre for Stratified Medicine (NICSM), Clinical Translational Research and Innovation Centre (C-TRIC), Altnagelvin Area Hospital, Ulster University. SNPs rs6551649, rs6551654, rs4495065, rs7940817, rs6724083, rs7596517, rs7593167, rs12911054, rs2569431, rs9844730, rs73590361, rs964680 are novel biomarker findings identified in this thesis and can possibly differentiate the groups, characterise and enable us to understand the treatment response and T2DM comorbid groups. The possible subgroups identified in this thesis using a clinical approach can aid enormously in the application of personalised medicine research. The genomic ML models built in this thesis with a classification accuracy of 83% can aid in early prediction of the comorbid T2DM groups. The clinical and proteomic ML models built in this thesis with a classification accuracy of 91% and 95% respectively can aid in identifying the subgroups of patients that would not respond to the treatment regime after diagnosis of T2DM through their circulating protein and bioclinical levels.

Inflammasomes are multi-protein complexes that serve as critical sensors to fend off pathogens and resolve aberrant cellular physiology. The NOD-, LRR-, and Pyrin domain-containing 3 (NLRP3) inflammasome can detect a wide range of stimuli, including pathogen-associated molecular patterns (PAMPs), damage-associated molecular patterns (DAMPs), and homeostasis-altering molecular processes (HAMPs). It drives the production of pro-inflammatory cytokine interleukin-1β (IL-1β) through the caspase-1 dependent signalling pathway. Despite extensive studies, no consensus model for the activation mechanism exists. Thus, systematic interrogation of NLRP3 inflammasome activation is highly desirable. Emerging mass spectrometry (MS) methodologies and proteomics have enabled the comprehensive mapping of intracellular protein networks. In this thesis, multiple MS-based strategies have been utilised to examine the global perturbation during NLRP3 inflammasome priming and activation. Using multiplexed TMT-based proteomics and di-Gly ubiquitomics, we presented a high-resolution map of proteome and ubiquitome profiles in THP-1 cells in response to NLRP3 activators. Through the combination of co-fractionation mass spectrometry (CF-MS) and machine learning algorithms, we elucidated the spatial and subcellular proteomic changes during the NLRP3 inflammasome activation process. We observe substantial relocalisation of the complex upon stimulation. Furthermore, utilising the APEX2 proximity labelling (PL)-MS approach, we spatiotemporally characterized the "proximity proteome (also known as proximitome)" of the NLRP3 inflammasome. Combined with RNAi screening, two NLRP3-interacting proteins, USP10 and UCHL1, were identified and validated as essential regulators for NLRP3 activation downstream of potassium efflux. Collectively, the MS-based approaches, including multiplexed proteomics, di-Gly ubiquitomics, CF-MS, and APEX2 PL-MS, enable a comprehensive understanding of the NLRP3 inflammasome activation process. These new insights into the underlying mechanism provide potential novel therapeutic targets, which may ultimately lead to treatments for NLRP3-associated diseases.

Prostate cancer (PCa) is the second most common type of cancer in men worldwide, with an estimated 1,414,259 new cases in 2020. The commonly used blood test for prostate specific antigen (PSA) has been used for the detection of PCa in its early stages. PSA screening, diagnosis, and prognosis have aided in the clinical management of patients with PCa. On the other hand, the PSA test has been criticised for the possibility of overdiagnosis and hence overtreatment of patients with indolent disease. Regardless of this, the PSA test continues to be the gold standard for PCa early detection. This argument highlights the unmet need for non-invasive PCa biomarkers with increased sensitivity and specificity capable of discriminating benign from malignant disease and predicting therapy response. The goal of this work was to identify prospective biomarkers in the peripheral blood of PCa patients using cutting-edge analytical techniques and machine learning. Sequential Window Acquisition of All Theoretical Mass Spectra (SWATH-MS) is a well-established data-independent acquisition (DIA) approach that requires prior knowledge of peptide mass spectrometry characteristics in order to extract relevant data on protein identity and quantification. By merging data from PCa patients with low, moderate, or high-grade prostate cancer and healthy controls, we were able to create a comprehensive PCa serum spectrum library containing thousands of peptides. Using the SWATH-MS method and this broad PCa spectral library, a 12-protein marker signature was identified. This implicated the complement/coagulation pathways and monocyte/macrophage development pathways in disease development. This signature was validated in two different patient cohorts and showed a constant ability in distinguishing between PCa patients and healthy individuals. Further, biomarker expression demonstrated a shift towards the levels seen in healthy individuals following treatment (radiation or prostatectomy). Additionally, nanoscale chromatography, which can take several hours per sample, is commonly used for proteome analysis in clinical research. In order to introduced automation and reduce run-time in proteomics research, we provided a high-throughput LC/MS/MS approach that utilises 1mm scale chromatography and requires around 15 minutes per sample. The reduced run time resulted in a 6-fold increase in productivity and excellent reproducibility when compared to nanoscale LC/MS/MS. Lastly, we took a multi-omics approach, another rapidly growing area of PCa research to develop improved biomarker panels (i.e., enhanced specificity) and contribute to a better understanding of PCa biology. In this study, we used a combination of lipid and protein measures and identified biomarker signatures that were capable of clearly distinguishing malignant from non-cancerous samples in our sample sets. Taken together, we propose a novel PCa blood-based spectral source and an automation approach for speedier proteomics sample processing, and we hope that they will be of use to researchers looking for PCa protein biomarkers in blood samples. Furthermore, to improve diagnosis in patients with localized and locally advanced PCa, our proteomics and multi-omics signatures would be helpful to facilitate diagnosis.

Background: The disease course of rheumatoid arthritis (RA) varies widely between patients, and various therapeutic options are available. However, none are universally effective, all have a risk of side-effects and currently, all are prescribed on a "trial and error" basis, based on escalating cost and not precision medicine targeted to patient endotype. Previous work has shown that circulating drug concentration levels of tumour necrosis factor inhibitors (TNFi, a class of drug used to treat RA and other autoimmune conditions) are associated with response to treatment. This thesis hypothesised that biological factors, such as protein expression, contribute to variability in circulating drug levels and treatment response to biologic agents in patients with RA. Methods: A population pharmacokinetic (popPK) study was carried out in patients with RA starting either Amgevita or Benepali, which are biosimilar agents for the TNFi agents adalimumab and etanercept, respectively. Model parameter estimates from the popPK study were used to simulate altered dosing intervals of these drugs. Proteomics data was obtained on all patients in the popPK study, as well as an additional cohort of patients with RA starting etanercept, using Sequential Window Acquisition of All Theoretical Mass Spectra (SWATH-MS). Protein expression was regressed against RA clinical outcome measures to determine any associations between protein expression and treatment response. Protein expression was also analysed alongside paired genotype data to determine whether any protein quantitative trait loci (pQTLs) existed. Significant pQTLs were then used to construct a polygenic risk score (PRS) for treatment non-response. Results: 16 patients were recruited to the popPK study; PK parameters were successfully estimated and used to simulate the effect of altered dosing intervals. SWATH-MS was used to generate proteomics data in serum samples from 180 selected patients commencing etanercept recruited to the Biologics in RA Genetics and Genomics Study Syndicate, a prospective multi-centre UK-based observational cohort. Proteomics analysis identified 52 proteins associated with RA clinical outcome measures. A pQTL analysis was carried out using 147 patients from the etanercept sub-cohort. 104 pQTLs were identified, 14 of which overlapped with significant proteins from the regression analysis. A PRS was generated using significant pQTLs, but was not found to be statistically significantly predictive of poor treatment response. Conclusions: The popPK study has provided proof-of-concept for future personalised dosing trials in patients with RA starting TNFi. This thesis has identified several proteins associated with RA clinical outcome measures that also have a genetic basis. Findings require external validation with replication studies in an independent cohort, but once confirmed, this could pave the way for future biomarker and/or drug target development.

Affecting over 250 million people worldwide, Osteoarthritis (OA) is the most prevalent musculoskeletal disease and the leading cause of disability among elderly people. An accelerated form of OA, termed post-traumatic OA (PTOA), is triggered by one or more injurious, high impact loading events which result in chondrocyte cell death and a 'phenotypic shift' in residual cells toward a more catabolic state. Conversely, cyclic loading within the physiological range has been shown to have beneficial effects on joint health, with longitudinal patient studies demonstrating a lack of rhythmic mechanical stimulation results in cartilage atrophy. Here, the differences in cartilage tissue response to physiological cyclic and pathological impact loads were examined, to elucidate the mechanosensitive pathways responsible for destabilising chondrocyte homeostasis, driving the initiation of PTOA. Assembly and optimisation of a novel ex vivo porcine loading model, incorporating an ElectroForce 5110 mechanical testing instrument, delivered compressive strains comparable to that seen in the joint, providing physiological cyclic (10% stain) and pathological impact (50% strain) loading regimes for further investigation. Biological readouts including cell death, sGAG release and structural damage demonstrated the catabolic and anti-catabolic effects of these two loading programmes. Subsequent development of a protein extraction protocol and employment of a data-driven discovery proteomics study permitted unbiased examination of the two aforementioned loading regimes and their short-term (< 72 h) global effects on cartilage health. Similarities were observed through a sub-population of indiscriminately mechanosensitive proteins associated with cytoskeletal remodelling, whilst divergence was detected in IL-1, integrin and mTOR signalling pathways, among others. Most notably, this work elucidated the induction of a number of small ubiquitin related modifier (SUMO)-associated proteins via mechanical stress ex vivo, chiefly following impact loading. Further investigation of this stress response pathway was conducted in a human chondrocyte cell line in vitro via co-immunoprecipitation mass spectrometry. A global shift towards deSUMOylation was observed following PIEZO1-mediated calcium influx, with the greatest alteration in protein SUMOylation state associated with RNA processing and the cytoskeleton. Overall, this study demonstrated the effective use of a novel ex vivo mechanical loading model and protein extraction methodology for the proteomic examination of the chondrocyte response to variable mechanical load. These techniques and the identification of a unique mechanosensitive pathway (SUMOylation) aim to further the field of orthopaedic research by progressing our understanding of PTOA disease progression in the hope of identifying new therapeutic targets for the future.

The human genome is involved in the development of neuropsychiatric disorders such as schizophrenia (SCZ), bipolar disorder (BD), and major depressive disorder (MDD), but the causal genomic molecular players remain unclear. Therefore, in this thesis, we evaluated disorder-associated disruptions within different genomic elements - genes, transcripts, novel open reading frames (nORFs: unannotated ORFs in the genome with recent evidence of coding potential), and their proteins - and considered how these elements contributed to disorder development. We used a combination of transcriptomic and proteomic methods, leveraging large RNA-Seq and microarray datasets to circumvent issues with power for the proposed analyses. A literature summary on previously identified transcriptomic and proteomic dysregulations in neuropsychiatric disorders, and motivations for the work presented in the thesis are introduced in Chapter 1. In Chapter 2, we evaluated sex-stratified cell count corrected case-control differences in gene expression and (where possible) transcript expression and transcript usage for five MDD RNA-Seq and microarray datasets derived from peripheral blood mononuclear cells (PBMCs) or whole blood. To improve detection of gene expression signals we used two meta-analytic approaches: a weighted Z-score and a Bayesian framework-based method, to compare 2,375 MDD cases and 3,606 CNT samples (largest meta-analysis to-date). Sex-specific dysregulations in gene and transcript expressions were identified. Pathway enrichment analysis showed that MDD was associated with dysregulations in innate and adaptive immunity, translation, complement cascade and chromatin organisation. Furthermore, theoretically motivated gene lists associated with MDD (e.g., cell cycle genes) were evaluated within each sex, and sex-specific dysregulations in replication-dependent histones, ribosomes, OXPHOS - electron transport chain, and nuclear mitochondrial genes were identified. In Chapter 3, we tested the hypothesis of MDD gene expression heterogeneity by identifying sex-stratified co-expressed clusters of genes present in five different MDD case-control studies and evaluating gene expression profiles in theoretically-motivated MDD subgroups. Weighted gene co-expression network analysis (WGCNA) of these datasets implicated gene modules related to response to wounding, nitrogen catabolism, and protein localisation as different between female MDD and CNT samples. MDD subgroups were identified using a data-driven approach, however, these gene expression-based subgroups did not show significant correlation with MDD associated phenotypic traits. We also evaluated previous evidence of an inflamed subgroup of MDD identified using cell count data and identified enrichments in immune-related pathways in both males and females. Furthermore, using C-reactive protein as a marker of inflammation status, we identified B cell signalling dysregulation in male cases of MDD compared to healthy controls. Therefore, MDD subgroups identified using phenotypic traits have distinct transcriptomic profiles, which should motivate future studies to perform subgroup-based MDD analyses. In Chapter 4, we explored the role of nORF and their protein products in neuropsychiatric disorders. nORFs - present within both protein-coding and noncoding regions of the human genome, represent an unexplored pool of genomic products within previously investigated genomic regions and could prove beneficial in understanding neuropsychiatric disorders. As the functional relevance of nORFs in the human genome is poorly understood, we conducted a systematic analysis to curate nORFs from three online sources, processed them extensively, and evaluated their genomic functions which revealed involvement of nORFs in nucleoside binding, protein-binding, and kinase-related processes. Furthermore, in comparing the genomic sequences of nORFs to canonical ORFs present within genes defined as protein-coding, we identified nORFs as shorter and more exon-rich sequences. Such evidence of noncanonical translations highlight the need to re-evaluate and update long standing binary categorisations of the genome into protein-coding and noncoding. In Chapter 5, a proteogenomic pipeline which combines proteomic and genomic datasets for nORF identification - proteomic data matches against genomic data to identify novel, unannotated peptide sequences such as nORFs, was described. However, proteomic datasets are not always available and some of the largest consortiums like GTEx (The Genotype-Tissue Expression project) and TCGA (The Cancer Genome Atlas) provide only RNA-Seq data. To mitigate this limitation, we developed a transcriptomic pipeline to identify nORFs within transcripts assembled from RNA-Seq data. The proteogenomic and transcriptomic pipelines were validated using datasets isolated from post-mortem human brains and mouse B and T cells and 3,054 transcribed nORFs and 1,658 translated nORFs were identified, respectively. In Chapter 6, the transcriptomic and proteogenomic pipelines we developed were used to identify nORFs in SCZ and BD samples. Commonly available RNA-Seq datasets are prepared using an rRNA depletion or poly(A) selection to remove highly abundant rRNAs which confound detection of transcripts with low abundances, or to select for mRNAs that contain poly(A) tails. As nORFs can be present within non-poly(A)-tailed transcripts, we required RNA-Seq datasets without poly(A) selection which were available for SCZ and BD but not MDD cases. We identified 3,103 transcribed nORFs of which 44 and 61 were differentially expressed in SCZ and BD, respectively. 21 translated nORFs were also identified. nORFs differentially expressed in SCZ and BD were found to be enriched for DHS1 (DNase1 hypersensitive sites) and histone modifications suggesting involvement in gene regulation. Two of the translated nORFs expressed in SCZ and BD but not CNT samples were present near interesting candidate genes: DISC1FP1 and STXBP1. Overlaps between transcribed nORFs, and SCZ- and BD-specific risk genes revealed nORFs to be enriched within SCZ- specific loci on chromosome 2. Functional predictions for the transcribed and translated nORFs revealed potential DNA-binding, enzymatic and cytoskeletal involvement. nORF expression was also found to be different between males and females, and subgroups of schizophrenic cases with and without psychosis. Therefore, nORFs are valuable genomic components that warrant further investigation in neuropsychiatric and other disorder contexts. In Chapter 7, the major findings from this thesis on the transcriptomic and proteomic dysregulations underlying neuropsychiatric disorders and how they add to our understanding of the human genome is discussed. Therefore, this thesis puts into broader clinical perspective what we know about the genetic architecture of neuropsychiatric disorders, including the novel genetic elements described in this thesis, and how this understanding could aid in the development of appropriate diagnostic and therapeutic strategies for these debilitating conditions in the future.

Nutraceutical modulation of lipotoxicity and the development of a tool to monitor palmitate induced sub-cellular dysfunction. - by Dean Frederick Ashley. Over 64.3% of the UK are currently overweight or obese according to the NHS survey, a number that has continued to increase throughout the beginning of the 21st century. Inherently this increases the use of NHS resources through an increased prevalence of obesity associated diseases, including type 2 diabetes. To address this, changes to individual's diets are required and a greater understanding of the sub-cellular mechanisms, that lead to insulin resistance and type 2 diabetes, need to be further understood. In this thesis, C2C12 myotubes & S. cerevisiae are treated with tunicamycin or the saturated fatty acid palmitate (Chapter 3), to create models of ER stress and lipotoxicity, respectively. These models are then used in subsequent chapters to study the nutraceutical modulation (Chapters 4 & 5) and sub-cellular aberrations of lipotoxicity (Chapter 6). Palmitate, the most prevalent saturated fatty acid elevated during obesity, was exposed to C2C12 myotubes in combination with the artificial sweetener Acesulfame-K (Ace-K)(Chapter 4) and the polyphenol compound curcumin (Chapter 5). Previous studies have highlighted a need for a more comprehensive understanding surrounding the possible effects of Ace-K. Using rtPCR and mass spectrometry, this thesis shows Ace-K can exacerbate palmitate induced inflammation, lipid intermediate accumulation and insulin resistance, through an AMPK dependant mechanism. Further consolidating the need for more in depth in vivo studies to understand the compound's risks. The same techniques are then used to study the effects of curcumin, in combination with palmitate, on C2C12 mouse skeletal muscle cells. Previous studies have shown curcumin alleviates palmitate induced inflammation, general lipid accumulation and insulin resistance. However, the specific lipid intermediate responsible for these changes remained unknown. This study demonstrates curcumin decreases diacylglycerol accumulation, through a βoxidative dependant mechanism, whilst simultaneously increasing long chain ceramide concentrations, through PERK signalling. This provides a possible combining mechanism for curcumins anti-inflammatory and cytotoxicity phenotypes. Lastly (Chapter 6), a protocol for monitoring the sub-cellular relocalisation of proteins and lipids is developed using the Localisation of organelle proteins by isotopic tagging (LOPIT) technique. Organelles are separated based on their densities and proteomic and lipidomic LCMS open-profiling is used to quantify proteins and lipids abundances across the differential centrifugation fractions. Known organelle marker proteins are then used by the Bayesian ANalysis of Differential Localisation Experiments (BANDLE) algorithm to assign organelle localisations to proteins and lipids with previously unknown organelle localisations, whilst simultaneously identifying species that have changed location following palmitate treatment. This offers a combined workflow to study the sub-cellular relocalisation of proteins and lipids that can be applied to a range of perturbations within C2C12 mouse skeletal muscle cells.

The post synaptic proteome is a complex of more than 3,000 proteins that form modular molecular machines that are involved in signal transduction, information processing and learning in the central nervous system. Previous analyses have used techniques from network science such as community detection to show differential enrichment of network modules using over-representation analysis for functions and disorders. In this thesis I extend these techniques by combining network analysis including community detection with Gene Set Analysis using the full polygenic signal available from summary GWA studies. Using large scale population discovery and replication samples I show that distinct communities in the post synaptic proteome have an enriched association for genes linked to human cognitive ability No evidence is found for an effect of network centrality on the propensity of genes to have variants associated with differences in intelligence but these are found to be associated with increased loss of function intolerance. This thesis describes a principled method in which network analysis of protein interactions can be used to understand human population genetic studies and could be extended to neuropsychiatric disorders affecting the synapse.

X-ray radiation is a common form of ionising radiation (IR) that naturally occurs at low levels in the atmosphere and Earth, which pose minimal threat to organisms as cells have multiple mechanisms to prevent damage or death. However, high levels of IR can overwhelm these mechanisms causing a cascade of events such as oxidative stress, DNA damage, cell cycle changes and cell death. The latter outcome is why IR has been used for radiotherapeutic treatments of highly proliferative cells in cancer for many decades. As well as this, it is often used as a method to induce DNA damage in research to study cell survival mechanisms. Targeting DNA damage repair (DDR) mechanisms in cancer cells has been of particular interest, as genomic instability is a key hallmark of cancers. Therefore, providing a therapeutic window for targeting cancerous tissue over healthy tissue. However, resistance to both radiotherapy and DDR-targeting therapeutics impose a big threat to the effectiveness of these onco-therapeutics. A lot of the recent advances in treatments of disease and understanding of resistance have been aided by omics research (i.e. genomic, transcriptomic and proteomic), which offer high-throughput measurements of cellular components and processes. Though, these methods mainly provide information of the changes in abundance or post-transcriptional modification status of molecules without the spatial context of where these molecules are located within the cell. Many proteins have multiple roles and these functions can be dependent on their subcellular localisation. More traditional methods for measuring subcellular localisation of proteins, such as microscopy are lower throughput, are biased/targeted and can be prone to artefacts. In this thesis, a subcellular proteomics technique, known as Localisation of Organellar Proteins using Isobaric Tagging with differential centrifugation (LOPIT-DC), was coupled with a semi-supervised Bayesian mixture model to address these technical limitations and screen for protein localisation events in an untargeted and cell-wide manner. First, protein expression was measured within the lung carcinoma A549 cell line over several time points, between 0 and 24 hrs, post-irradiation with x-rays using quantitative proteomics, and 2 and 12 hr time points were subsequently selected for further investigation. As phosphorylation is one of the most abundant post-translational modification and drives a lot of protein trafficking events, quantitative phosphoproteomics was also performed. Then, LOPIT-DC used to assess the immediate and delayed protein trafficking responses at 2 and 12 hrs post-IR. Using this approach, 504 and 829 proteins, respectively, were found to differentially localise upon IR with associated biological pathways that included cell cycle checkpoints, transcription by RNA polymerase II, immune system and iron metabolism. Within the 12hr timepoint, several proteins involved in the ferroptosis pathway, a type of programmed iron-dependent cell death, had changed localisation, 3 including a key ferroptosis suppressor. Immunofluorescence microscopy was used to validate several of these differentially localised ferroptosis proteins, AIFM2, TFRC, FTH1 and FTL. Collectively, the subcellar changes of these proteins indicated a potential suppression of ferroptosis, thus a radioresistance mechanism. However, this would require further validation with functional assays. According to our orthogonal proteomics data, changes in subcellular localisation of the proteins appeared to be relatively independent of protein expression or phosphorylation state, indicating that these potentially key cellular responses may be overlooked in other assays. These high-content datasets provide a powerful resource for further investigating the functional significance of such x-ray-dependent differential localisations, potentially finding novel targets associated with oxidative stress and DNA repair that play key roles in disease and radioresistance.

Multiple sclerosis (MS) is a chronic inflammatory demyelinating disease of the central nervous system where neurodegeneration often leads to irreversible disability. Clinical and pathological heterogeneity is the rule, with genetic (e.g. HLA-DRB1*15 status) and environmental factors playing important roles. The mechanisms underlying such heterogeneity remain elusive. Astrocytes, which are involved in MS pathogenesis, are an attractive candidate to contribute to such variation in MS pathology due to their complex functions in health and disease. Recent progress has also challenged the conventional view that astrocytes represent a homogenous population, indicating that astroglial diversity may underscore MS heterogeneity. In this thesis, we undertake a quantitative neuropathological analysis of GFAP+ and ALDH1L1+ astrocyte subpopulations in the motor cortex of a large cohort of human post-mortem MS cases genotyped at the HLA-DRB1*15 locus. We find that GFAP+ and ALDH1L1+ astrocytes represent distinct subpopulations that differ in their distribution along the motor cortex and in their relationships with MS pathological outcomes. GFAP+ astrocytes show a greater extent of reactivity in the MS motor cortex, especially in areas of demyelination and in younger HLA- DRB1*15+ cases. In contrast, MS ALDH1L1+ astrocytes associate with neuronal density and macrophage/microglial inflammation, a relationship not seen in cases carrying the HLA-DRB1*15 allele. These astrocyte subpopulations also vary in their relationship with surrogate markers of blood-brain barrier dysfunction with decreasing perivascular GFAP+ astrocyte expression and increasing perivascular ALDH1L1+ astrocyte expression in MS. Given the histopathological differences between the two astrocyte markers, we also perform proteomics using laser capture microdissection coupled to mass spectrometry to probe the deep proteome of GFAP+ and ALDH1L1+ astrocyte subpopulations. Protein expressions show clear differences between GFAP+ and ALDH1L1+ astrocytes. In MS cases, GFAP+ astrocytes abundantly express immune-related proteins compared to ALDH1L1+ astrocytes, which associate with protein expression changes related to synaptic vesicle clustering and voltage-gated potassium channels. Immune-related proteins in astrocyte subpopulations are impacted by the carriage of the HLA-DRB1*15 allele. Altogether, our data implicate astrocyte subpopulation diversity as a contributing factor of MS heterogeneity. We suggest that GFAP+ astrocytes have relatively detrimental roles, whilst ALDH1L1+ astrocytes have relatively protective roles in MS. This thesis highlights the pressing need to understand the role of astrocyte diversity in MS so that selective targets amenable to therapeutic modulation may be identified and halt the devastating outcomes of MS.

The endoplasmic reticulum (ER) is the largest organelle in the cell and is involved in a wide range of functions such as biogenesis of membrane and secretory proteins, calcium homeostasis and lipid synthesis. To be able to carry out these functions, the ER depends on an elaborate architecture consisting of a network of flat cisternae-like sheets and narrow tubules. The ER also assembles the nuclear envelope (NE) which is formed by two membranes, the inner and outer nuclear membrane (INM and ONM), that separate the cytoplasm from the nucleoplasm. Despite continuous with each other, these ER domains serve different functions. Consistent with their diverse roles, ER domains also appear to harbour distinct proteins. However, the complete proteomes of the various ER domains have not been systematically analysed and annotated. The primary objective of this project was to map the protein compositions of the distinct ER domains. To this aim, an APEX2-based proximity biotinylation assay coupled to subcellular fractionation was established and a data analysis pipeline developed. The assay was capable to identify many of the well-characterised INM, ER tubule and ER sheet proteins in their anticipated domain. Additionally, a range of bona fide proteins, that have not previously been shown to enrich at any of these domains, were identified. Therefore, these data provide a valuable resource to enhance our understanding of ER organisation. Additionally, the developed assay and data analysis pipeline can serve as a discovery platform to probe how the protein compositions of the ER domains change in response to disease-associated mutations or upon external perturbances. As an example, the effect of LMNA knockout on the protein composition of the inner nuclear membrane was investigated, whereby a small number of proteins were identified to be affected by depletion of A-type lamins.

Introduction: Pancreatic ductal adenocarcinoma (PDAC) has very low survival rates that have not changed in over 40 years. Delay in diagnosis and the development of drug-resistance limits the effectiveness of current therapies, making identification of novel early detection and combination treatment strategies, extremely important in order to improve survival rates. Materials and Methods: An in vitro model of gemcitabine and 5FU resistant PDAC cell lines were generated in the lab by continuous treatment with the drugs. Reverse Phase Protein Array (RPPA) and Liquid Chromatography-Mass Spectrometry (LC-MS) based analysis of protein expression and metabolite levels was carried out in PDAC cell line models on chemotherapy treatment in order to identify upregulated proteins and metabolites. Based on the importance of targeting downstream signalling pathways to KRAS in PDAC and the joint proteomics and metabolomics analysis, AKT was chosen as a target that could potentially enhance chemotherapy effectiveness in PDAC. AKT inhibition was assessed in combination with single chemotherapy agents. Based on these findings and previous research in our lab in Oesophageal Adenocarcinoma (OAC) cell lines, AKT inhibition in combination with anti-apoptotic protein inhibition was assessed. Cell viability was assessed using MTT assays and CI values were generated to assess synergy. Proliferation assays, western blotting, flow cytometry and high content analysis were used to decipher mechanism of action of the combination treatments. Chemoresistant PDAC cell lines were used to assess AKT's role in acquired resistance and to investigate the efficacy of combined AKT/IAP inhibition. Additionally, to identify the biological pathways implicated in the early development of PDAC, a systematic review was carried out in order to assess the association between protein and metabolite based biomarkers and PDAC risk (published findings included manuscript in final appendix, Chapter 9). Results: Protein expression analysis using RPPA showed increased expression of several receptor tyrosine kinases (RTKs) and a few downstream mediators to KRAS. In addition, the top dysregulated proteins based on fold change were identified on combination chemotherapy treatment that could serve as potential interesting combination targets. Jagged-1 and Connexin-43 were also found to be highly upregulated in the chemoresistant lines, indicating a role in governing chemoresistance. LC-MS based metabolomics analysis showed upregulation of TCA cycle metabolites and non-essential amino acids in the PDAC cell lines. Links to AKT signalling were identified with the RTKs identified, EGFR, EphA2, IGF all potentially acting through AKT signalling and an extensive literature review also highlighted the potential role of this pathway, downstream of KRAS in regulating the metabolic changes seen. Based on these observations and previous work in our lab in OAC cell lines assessing AKT inhibition, the efficacy of AKT inhibition was assessed. Two allosteric AKT inhibitors were found to synergize with the chemotherapy drugs oxaliplatin, gemcitabine and 5FU but the combination was found to induce only cytostatic/anti-proliferative effects. In order to swing the balance towards cell death, AKT inhibition was assessed in combination with anti-apoptotic proteins and found to be highly synergistic. Mechanism of action studies revealed that both a caspase dependent and independent mechanism of cell death contributed to this synergy. Interestingly, this combination was also effective in reducing cell viability of in vitro chemoresistant lines generated in the lab. Additionally, our pooled analysis in the form of a systematic review demonstrates that higher serum glucose levels and lower levels of PLP are associated with risk of PDAC and highlights the importance of more prospective studies that are required to identify biomarkers for early detection (published findings included manuscript in final appendix, Chapter 9). Conclusion: A combined proteomics and metabolomics based approach was used to identify novel treatment strategies for PDAC and identified that combined inhibition of AKT and IAPs could synergistically reduce cell viability in parental and chemo-resistant PDAC cell lines. This suggests that this combination could be used as a second line therapy in chemo refractory PDAC. Our results suggest that the dual targeting of AKT and IAPs could be an effective therapeutic combination in PDAC.

The deep sea is a vast, remote, and relatively unexplored environment that is increasingly viewed as an opportunity for industrial exploitation. Scientists have been calling for the conservation of the deep sea for decades, to preserve the unique benthic habitats and species that thrive in this extreme environment. Conservation efforts continue to be impeded by a severe lack of baseline data such that current records on deep-sea habitats and their associated faunas are inadequate to inform appropriate management and conservation strategies. As anthropogenic threats escalate at an unprecedented rate, the balance between the conflicting forces of conservation need and lack of baseline data increasingly require a precautionary approach. This thesis therefore aimed to combat this conservation impediment, first by addressing the need for improved fundamental knowledge about the systematics, distributions, and relationships of deep-sea benthic invertebrates by contributing new data for understudied taxa, and then by exploring how deep-sea conservation may be achieved with the limited data that are already available. I used phylogenetic analyses, including novel sequence data for holothurians, to better understand species identity and distributions at deep-sea benthic habitats. This understudied group provided a case study to examine how issues of species connectivity, whether locally endemic or globally widespread, are crucial for developing effective conservation measures. Deep-sea holothurians of the genera Elpidia Théel, 1882 and Chiridota Eschscholtz, 1829 were found to have minimal intraspecific genetic variation, despite having widespread geographic distributions, suggestive of sustained connectivity even across environmental barriers. Using deep-sea gastropods, I further established a proof of principle for the application of proteomic analyses as an emerging molecular tool to explore the physiological relationships that underpin the ecology of species at chemosynthetic environments. This provided an important foundation to further our understanding of the complex interactions at insular deep-sea habitats. These analyses suggested that the Antarctic vent gastropod Gigantopelta chessoia C. Chen, Linse, Roterman, Copley & Rogers, 2015 hosts its sulfur-oxidising bacterial endosymbionts exclusively within the specialised trophosome organ, as opposed to the more commonly observed gill-based symbioses, with no evidence of a dual symbiosis. Finally, given the advancing development of the deep-sea mining industry, I demonstrated how a universally known conservation tool, the IUCN Red List of Threatened Species, can be robustly applied to even the most understudied deep-sea taxa in a habitat entirely new to the red-listing process. Using molluscs, I provided the first global assessment of extinction risk for an entire taxonomic group at hydrothermal vents, exemplifying how taxonomy-driven tools can be utilised to support deep-sea conservation and providing a precedent for the application of the Red List across diverse deep-sea species and habitats. This revealed that 62% of all known vent-endemic mollusc species are currently threatened by deep-sea mining and that seabed management and mining regulation have the greatest impact on these extinction risk assessment outcomes. Given the complexities and uncertainties surrounding deep-sea conservation, this thesis summarises how we can utilise and build on existing data to ensure the conservation of deep-sea benthic invertebrates.

Pulmonary exacerbations (PEx) are clinically impactful events that accelerate Cystic Fibrosis (CF) lung disease progression. Current PEx treatment approaches have suboptimal outcomes and the frequency of PEx experienced by people with CF (pwCF) is variable and poorly predicted by demographic or clinical factors. There is an unmet need to improve the prediction, treatment, and management of PEx, however this has been hampered by a lack of understanding of the pathophysiology of PEx. Characterisation of the underlying host immune and microbiological mechanisms underlying PEx could improve therapeutic management of PEx and inform a more individualised approach to treatment. In this thesis we provide novel insights into the pathophysiological mechanisms underlying infrequent and frequent exacerbations which we anticipate will be highly informative for improving the personalisation of treatment of PEx through targeting the underlying disease processes.

Down Syndrome (DS), also called trisomy 21, is the most common non-lethal fetal aneuploidy that affects 1 in 800 live births. The disease appears mostly due to the existence of an extra copy of chromosome 21. In the UK, the screening of DS is offered to all pregnant women in the antenatal care program to assess the risk of DS. If the risk is high the women are offered further testing that includes invasive sampling (amniocentesis or chorionic villus sampling) to confirm the diagnosis. This project aimed to investigate the possibility of applying new methods to screen for DS which will give women at a medium risk of having a DS baby the chance to avoid unnecessary invasive interventions. These methods include liquid chromatography mass spectrometry (LCMS) to identify possible new biomarkers for DS pregnancies and investigate the potential use of PRSS56 and SPINK7 proteins as two new target biomarkers for DS. We will also use multiplex droplet digital PCR (ddPCR) using cell-free fetal DNA (cffDNA) to target different genes on chromosomes 21, 18, and 13 to screen for trisomy 21 (T21), trisomy 18 (T18) and trisomy 13 (T13), respectively. PRSS56 and SPINK7 proteins are secreted proteins and found to be highly expressed in the DS fetus placenta. The level of these proteins was investigated using depleted plasma samples from DS pregnancies and compared to plasma samples from healthy individuals (controls). The plasma samples were subjected to immunoprecipitation and Western blotting using monoclonal and polyclonal antibodies. PRSS56 was not detected in the plasma samples from DS pregnancies. SPINK7 was difficult to detect due to its small size and furthermore optimisation is needed. A group of 14 depleted plasma samples from DS and healthy control pregnancies were tested by LCMS to evaluate the abundance of the proteins in each group. The differential analysis showed 47 proteins were abundant in DS pregnancies with more than 2-fold change compared to control samples. Following the presence/absence analysis on the mass spectrometry dataset, the actin-related protein 2/3 complex subunit 2 (ARPC2) was detected in controls and absent in most of the DS pregnancies. ddPCR was investigated as a possible non-invasive prenatal testing (NIPT) platform to assess its ability and efficiency to analyse cffDNA. Samples were spiked by trisomy genomic DNA for T21, T18 and T13 and were tested to mimic the low fraction of cffDNA in the maternal blood samples. The T21 multiplexing assay targeted HSPA13 and VPS26C genes located on chromosome 21 against LEPROT and AGO1 genes located on on chromosome 1 as reference genes. The T21, T18 and T13 multiplex assay targeted VPS26C gene located on chromosome 21, BRCA2 gene located on chromosome 13 and LPIN2 gene located on chromosome 18 against AGO1 gene located on chromosome 1 as a reference. After the optimisation, 16 clusters were observed and each cluster was isolated with minimum raindrops, where the threshold lines can be easily placed between clusters. In conclusion, finding new biomarkers for DS will improve the detection rate of screening assays, which will reduce the need for invasive interventions. Studying protein abundance in DS pregnancies and different pathways will help us to find new biomarkers and understand their potential involvement in the severity of the disease. NIPT requires a rapid, sensitive and simple platform that is capable to test cffDNA efficiently. ddPCR has the potential to replace next-generation sequencing and microarrays because it is cheaper and easier to perform on a daily basis routine.

Acute coronary syndrome (ACS) causes the majority of deaths in ischaemic heart disease. Hence, ACS needs to be rapidly and correctly diagnosed. The electrocardiogram (ECG) is a non-invasive method that is used to diagnose ACSs alongside cardiac biomarkers such as troponin. However, ACS diagnostics when using the ECG and blood biomarkers have several challenges such as poor ECG data quality due to electrode misplacement or noise. Moreover, there are limitations with the use and accuracy of cardiac biomarkers since they typically require a waiting time to attain the result. The aim of this PhD is to use data science to augment clinical decision-making by improving, 1) ECG data quality and 2) discovering novel biomarkers to improve the sensitivity and specificity of detecting acute myocardial infarctions, specifically ST-elevation myocardial infarction (STEMI). Datasets including blood biomarkers and ECGs have been collected from Altnagelvin hospital. In the first part of this PhD, several traditional machine learning (ML) and deep learning (DL) algorithms have been used to detect electrode misplacement in ECG data. The ML algorithms achieved high performance rates (accuracy = 93% and 0.9 area under the curve). In the second part of this PhD, the impact of signal noise on ECG interpretation was investigated, and it transpired that noise has a significant impact (P < 0.001 McNemar's) on ECG classification when using ML. In the third part of the PhD, ML has been used to discover novel blood biomarkers that could be used to discriminate between STEMI, Non-STEMI and unstable angina. In this part of the PhD, we investigated whether combining those detected blood biomarkers with ECG features can improve STEMI detection. According to the results from each part in this PhD, the use of ML demonstrated promising results when applied to these cardiology datasets. This PhD could be used to inform the development of a new medical instrument or medical laboratory kits that can provide timely and accurate diagnoses.

Background: Microvascular complications are common in diabetes mellitus (DM). Approximately 30% of the patients with DM developed chronic kidney disease (CKD), also known as diabetic kidney disease (DKD). DKD is associated with higher mortality risk before progressing to kidney failure. The integration of electronic health records (EHRs) and high-throughput multiomics data has shown great potential in personalised medicine. This thesis aims to identify biomarker clusters (corresponding to underlying systemic factors) of CKD progression in type 2 diabetes (T2D) using EHR-linked Scottish cohorts (GoDARTS, SUMMIT, and RHAPSODY). Methods: Data included routinely-measured biochemistry variables and prescribing information in EHR; lifestyle factors and DM-related risk factors in GoDARTS; proteomics and metabolomics data of stage 3 CKD in SUMMIT; and proteomics, metabolomics, and lipidomics data of early CKD in RHAPSODY. Kidney function trajectory was estimated based on serum creatinine levels. Standard statistical models, dimensionality reduction methods, and functional enrichment analyses were applied. Results: We identified three main biomarker clusters corresponding to albuminuria, DM, and cardiovascular disease (CVD). Other less significant biomarker clusters could be associated with a variety of stage-specific symptoms such as diabetic retinopathy (DR), smoking status, potassium level, infection, and mucous membrane inflammation. In addition, the functional analysis highlighted the importance of inflammation, endothelial dysfunction, and lipid metabolism. Conclusions: The biomarker clusters improved the prediction accuracy by incorporating both strong and weak signals from individual biomarkers. The clusters could also serve as a framework for interpreting the underlying pathophysiology. The holistic approach identified a comprehensive set of prescriptions and biomarkers, which warrant further investigations.

Microsporidiosis poses an ongoing global threat to human health and animal wellbeing, in immunocompromised populations. As obligate intracellular parasites, microsporidia possess a greatly reduced genome and metabolome, whilst exhibiting an extensive array of extracellular proteins for host invasion and adherence. The aim of the work described in this thesis is to provide a clearer understanding of the microsporidian infectious pathway and presents a three- step pipeline to aid the discovery of novel proteins and mechanisms within that infectious pathway. Throughout the study, a combination of in-silico and laboratory-based investigations were performed to place equal emphasis on the proteins of the parasite and their respective hosts. Firstly, an in-silico experiment was devised using hidden Markov modelling and orthology-based prediction to collate and expand current putative microsporidian spore wall and polar tube proteins. Next, a novel proteomic "SHIPP" method was designed to isolate and identify novel microsporidian surface proteins and their respective host target sites. Lastly, two transfection methods of microsporidia were attempted to validate spore-host interactions: namely, trial use of cell penetrating peptides; and an adaptation of the CHoP-In mutagenesis approach for the knockdown of host membrane proteins. The results produced a consensus database comprising 294 surface protein orthologues, split across 34 distinct protein types. As such it represents one of the most comprehensive databases of its kind, to date. A trial of the "SHIPP" assay was able to successfully select for the extracellular proteins of both microsporidia and host cells, whilst identifying six novel protein candidates in microsporidia for further characterisation. Finally, use of cell penetrating peptides was shown to be an ineffective method for microsporidian transfection, whilst the adaption of the CHoP-In protocol for the mutagenesis of host cells requires further work. CHoP-In components were successfully designed, but could not be fully implemented into a complete trial run. Ultimately, it is believed that the components of this pipeline will assist in the discovery of novel microsporidian surface proteins, with future adaptations potentially assisting with the design of novel microsporidian treatments and the generation of resistant livestock to minimise the impacts of microsporidiosis.

Standard parametric statistical approaches based on comparison to global activity tend to perform poorly when this activity varies over multiple scales. Such multiscale variation, termed long range dependence, is a well-documented features of many biological and neurological data sets. We provide evidence from the literature as well as from data that demonstrates long range dependence across three contexts in: protein, brain and neuronal data. We propose novel non-parametric statistical approaches that account for these dependencies using network models. In networks of annotated proteins (nodes) connected by their physical interactions (edges), communities of functionally related proteins can provide possible novel drug targets. However, this guilt by association approach is vulnerable to potential bias in protein annotations. By restricted random-walk-based ranking of communities, our method, CommFinedWalker, preserves protein co-expression within well-ranked communities (suggesting functional conservation). The method provides novel drug target candidates that add to the diversity of interesting proteins and potential drug targets for research. As proof of concept, we explore a case study of a community that contains proteins involved in a rare genetic disorder. These results also serve as inspiration for novel methods to rank interesting communities in other contexts including brain state networks. Brain states have previously been studied across many conditions, including sleep and disorders of consciousness using a variety of models. However, brain state has not yet been studied in pharmacologically-induced unconsciousness. We define a new subclass of models, Hidden Markov Graph Models, to study brain states and communities of brain regions in fMRI data from subjects in both wakefulness and anaesthesia-induced unconsciousness. The framework draws on principles of free energy minimisation and entropy maximisation. We show evidence of characteristic resting state network activity in wakefulness as well as states of reduced temporal and functional complexity that are unique to anaesthesia-induced unconsciousness. These states further relate to established EEG markers of unconsciousness and long range dependent slow waves with implications for effective anaesthesia concentration in surgery. Lastly, we propose a pipeline to guide experimentation and perform automated detection of synaptic calcium events from single neuron light sheet microscopy data. This pipeline iteratively combines experimentation with experimentally informed simulation to improve experimental outcomes. The model fitting framework we provide draws on the free energy principle, and detection and smoothing methods that account for long range dependence and non-stationary activity present in the microscopy background. Through simulated results we empirically show that our methods can provide consistent estimates of actual experimental conditions that could be used to inform future experimentation.

There have been rapid advances in sequencing and proteomic technologies over the last 20 years. However, gaps remain, both in terms of genome availability for many species, and in the use of these technologies in different fields. The field of comparative immunology is steadily advancing, but genetic data from an even wider range of animals is required before we can fully understand the evolution and function of immune genes and systems. High-throughput proteomic analyses are now relatively common in mammalian studies but also offer tremendous potential to the field of comparative immunology. The work in this Thesis combines such analyses with available genomic data for a teleost and agnathan, and with newly generated transcriptomic data for a cartilaginous fish. It demonstrates that a wide range of different questions can be addressed using this approach, and that the large amounts of novel data are obtained with huge potential for follow-up studies, such as protein characterization and functional studies. Here, I examined post-immunization changes in the plasma proteomes of several distantly related vertebrate species. The plasma is an excellent indicator of immunological and broader biological status and can be studied using non-lethal blood sampling. This included assessment of the individual level response of a teleost fish (rainbow trout Oncorhynchus mykiss) and cartilaginous fish (nurse shark Ginglymostoma cirratum) to immunization using repeat sampling. The results have implications for the future assessment of immunological responses and add greatly to our knowledge of plasma proteins and their immune regulation in both species. I also compared the plasma proteomes of the above species to each other and to a jawless fish (sea lamprey Petromyzon marinus), enhancing our knowledge of the evolution of immune protection and functional convergence. The data generated, along with recently available genomic resources, enabled a timely reassessment of two longstanding questions about the evolution of key plasma proteins (serum albumin and the complement system), demonstrating how proteomic data can enhance phylogenetic studies. Overall, this work has boosted our comparative knowledge of the biology and immune function of distantly related vertebrate species and greatly enhanced both the available genetic resources for the nurse shark and the proteomic data for all species studied.

Background: Cirrhosis is a chronic illness that reduces liver functions including drug metabolism. Many drugs that are available in the market lack dosage guidance for hepatic impairment patients. Including those patients in the early phases of clinical trials can be risky if safe doses were not used. Whilst precision medicine is not a new concept, optimising in silico modelling such as physiologically-based pharmacokinetic (PBPK) modelling and simulation (M&S), can assist in informing drug labelling. This area is quickly growing over the last years, especially for special patient populations. Methods: Literature reviews were performed to understand the current situation with drug dosing in hepatic impairment, the quality of current PBPK models, and identify the gaps. Experimentally, scaling factors for converting clearance data from in vitro to in vivo in cirrhosis were determined. LC-MS/MS-based targeted proteomics were implemented to quantify the abundances of 51 drug-metabolising enzymes and transporters (DME&T) in 32 liver samples from cirrhosis patients at different degrees of disease severity compared to 14 normal control samples. Results: Microsomal and cytosolic protein contents decreased in cirrhosis relative to control samples and varied according to associated liver pathologies. Disease perturbation factor (DPF) reconciled differences in absolute abundances between various proteomic data analysis methods. Specifically designed heavy-labelled concatenated unique peptides from target proteins showed good performances as internal standards with the samples. Abundances of most DME&T per gram liver were lower by 30-50% in mild, 40-70% in moderate, and 50-98% in severe cirrhosis groups compared to controls. DPF, used as a scalar for protein abundances in PBPK models for repaglinide, dabigatran etexilate, and zidovudine, helped to enhance models' predictive performance. Conclusion: This thesis, to our knowledge, provides the first comprehensive quantification of relevant DME&T in all stages of liver cirrhosis. This helps the development of existing in silico cirrhosis models to inform drug labelling and recommends dose adjustments in scenarios that have not been studied clinically.

Biological tissues contain diverse cell subpopulations crucial to their function, and proteins are the macromolecules that carry out many of these functions. Robust mass spectrometry-based methods that can accurately identify and quantify thousands of proteins within a sample have transformed the investigation of protein abundance and function within tissues and cells. However, mass spectrometry-based proteomics of tissue is traditionally performed on bulk samples. Therefore, information on the spatial composition and protein abundance patterns of those tissues is lost. Currently, the spatial context of proteins within a tissue is investigated using antibody-based high-resolution techniques with little multiplexing capacity. There is a clear need for techniques that generate unbiased, quantitative, spatial expression profiles of many hundreds to thousands of proteins within a tissue. Presented within this thesis is the development of sample preparation, acquisition, and analysis workflows for spatially resolved proteomics using bottom-up mass spectrometry. This method was used to analyse an atypical teratoid/rhabdoid tumour (ATRT), a rare paediatric brain tumour. Firstly, an optimised workflow was developed and used to characterise the proteomes of two neurons: Betz and Purkinje cells. Both cell types show highly specific proteomes. Purkinje cells show an increased abundance of several known Purkinje cell marker proteins. In addition, several novel candidate marker proteins for Betz cells were identified. Secondly, the method was extended to allow for unbiased sampling of tissue in a spatially resolved manner. The method shows clear signals from a starting material of approximately 50 cells isolated from ATRT tissue. Analysis of the data within the tissue's spatial context highlights proteins and pathways showing spatially variable abundance within the tumour and adjacent normal tissue. This is performed without prior knowledge of the tissue composition by using spatially aware statistical methods. Finally, orthogonal validation was performed using immunohistochemistry and mass spectrometry imaging which recapitulates the spatial variability observed in the bottom-up mass spectrometry approach. Overall, the work presented here informs on methods for spatially resolved proteomics of tissue, laying the groundwork for further spatially resolved tissue proteomics and highlighting the utility of such approaches to understanding tissue biology.

Witnessing a decade since the Physics Nobel Prize awarding for the ground breaking studies on the 'two-dimensional material graphene', graphene oxide (GO) stands as a promising candidate under the spotlight of biomedical research. Amongst the most questioned applications lies the possibility of its future use for intravenous (IV) treatment, imaging, diagnostic or theranostic purposes. Numerous studies are highlighting the trade-off between the physicochemical properties of GO materials and their in vivo safety, biodistribution and performances. However, little is known about the in vivo interaction of GO with biofluids. To date, no study investigated the interactions of GO with proteins upon their IV administration. This is a major gap, as it has been shown that the adsorption of blood proteins on the surface of nanoparticles can modulate their in vivo effects and fate. Our group and others have found that GO flakes are able to travel from blood to bladder, without causing renal impairment. In this thesis, we show that renal excretion could be explored to recover protein-coated GO flakes after their IV administration. We developed a robust method to simultaneously recover GO and remove free proteins from urine. Our results suggest that the proteomic content of mouse urine is enriched upon GO excretion. Further on, inspired by the proteomic studies in our laboratory showing that in vivo protein-coated nanoparticles can be explored as a disease biomarker resource, we questioned the possibility of using our method for scavenging disease-related proteins. Using B16-F10 melanoma tumour bearing mice, we provided evidence that the excretion of GO boosts the levels of disease-relevant molecules into the urine samples. In summary, this project reveals a new method for the in vivo GO-protein interaction studies and disease biomarker scavenging.

Neuronal communication occurs at specialised sites, known as synapses. Presynaptic neurotransmitter release from one neuron acts on the postsynaptic receptors of another, facilitating transfer of electrical impulses between neurons. At glutamatergic synapses, ionotropic glutamate receptors (iGluRs) respond to presynaptically released L-glutamate, the predominant excitatory neurotransmitter. Excitatory synaptic transmission is mediated primarily by AMPA-type glutamate receptors (AMPARs), and is crucial for the transfer and storage of information in the brain. The AMPAR engages in numerous protein interactions, modulating receptor trafficking, channel gating, synaptic positioning, and ultimately synaptic strength. iGluR family members bind synaptic proteins through their N-terminal domain (NTD), influencing synaptic and neuronal circuit function. The AMPAR NTD encompasses 50% of the receptor, extending midway into the synaptic cleft, exposing it to a protein-rich environment. The strength of synaptic transmission is dependent on the AMPAR NTD, likely by positioning receptors at the synapse. This mechanism is thought to be regulated through direct protein interactions with the NTD. The identity and functional importance of these protein interactions however, are unknown. Therefore, this study aims to identify and characterise the synaptic function of AMPAR NTD-interacting proteins. Aided by the development of a novel proximity-labelling proteomics technique, synaptic proteins enriched for the AMPAR NTD were determined. This permitted identification of transient NTD interactors, previously unresolved using classical affinity-purifications. Furthermore, this technique provides a comprehensive list of diverse synaptic cleft proteins, which may influence AMPAR function through direct or indirect mechanisms. Candidate interactors were later screened for direct interactions with iGluR NTDs by establishing a cell-based binding assay. This confirmed known iGluR NTD interactors and uncovered neuronal pentraxin-1 (Nptx1) as an AMPAR NTD-interacting protein. Facilitated by structural biology, the Nptx1 binding site was deduced using AMPAR NTD mutants. The functional consequence of this interaction was then studied at synapses onto principle neurons of the hippocampus using different electrophysiological and imaging approaches, demonstrating Nptx1 as a subunit-specific regulator of AMPAR-mediated synaptic transmission. These findings have profound implications for understanding the role of AMPAR-interacting proteins in regulating the strength of synaptic connections.

The global burden of age-related cognitive decline and dementia will continue to rise in tandem with our ageing population. This necessitates the discovery of novel biomarkers and candidate drug targets to combat cognitive dysfunction. Blood proteins are important drug targets, and blood samples can be acquired routinely in clinical settings and epidemiological studies. Whereas hundreds of blood proteins are associated with cognitive ability and dementia, we do not understand whether these associations represent correlation or causation. Genome-wide association studies (GWAS) are required to define variants that are associated with blood protein levels. These variants can proxy for candidate disease-markers and assess their causal associations with health outcomes in analysis methods such as Mendelian randomisation. DNA methylation is an epigenetic mechanism that regulates gene expression and is influenced by genetic and environmental factors. Studying the relationship between DNA methylation and protein levels could reveal whether genetic variation or environmental factors likely mediate associations between blood proteins and disease states. The first aim of this thesis is to conduct GWAS and epigenome-wide association studies (EWAS, using DNA methylation) on plasma levels of 422 unique proteins. Using these data, I apply causal inference approaches to determine whether blood proteins are causally associated with Alzheimer's disease risk. Several strategies have been proposed to estimate biological age by leveraging inter-individual variation in DNA methylation profiles. Epigenetic measures of ageing correlate strongly with chronological age. Recently, a novel epigenetic measure of ageing termed 'DNAm GrimAge' was developed to predict one's risk of mortality. DNAm GrimAge is a composite biomarker that incorporates methylation-based predictors of seven blood protein levels and smoking. The relationship between this biomarker of ageing and cognitive decline or dementia is not known. Therefore, the second aim of this thesis is to examine whether DNAm GrimAge associates with measures of brain health and Alzheimer's disease. To conduct these aims, I utilise data from two cohort studies: the Lothian Birth Cohort 1936 (n ≤ 906, LBC1936) and Generation Scotland (n ≤ 9,537, GS). In Chapters 1-3, I provide an overview of cognitive ageing and dementia. I describe GWAS and EWAS on blood protein levels and the development of DNAm GrimAge. In Chapter 4, I detail the population cohorts and main methodologies that are used in this thesis. In Chapter 5, I conduct GWAS and EWAS on plasma levels of 92 neurology-related proteins (n ≤ 750, LBC1936). I identified 41 independent genetic and 26 epigenetic loci that associate with 33 and 9 proteins, respectively. I showed that an immune-related protein, poliovirus receptor (PVR), is causally associated with Alzheimer's disease risk. In Chapter 6, I use a novel Bayesian framework termed BayesR+ to perform an integrated GWAS/EWAS on plasma levels of 70 inflammation-associated proteins (n = 876, LBC1936). Many GWAS and EWAS use linear models, which examine every measured genetic or epigenetic site in isolation. BayesR+ accounts for intercorrelations among genetic and epigenetic sites and the reciprocal influences of these data types. I estimated the contribution of genetic and epigenetic variation towards inter-individual differences in inflammatory protein levels, considered alone and together. There was no evidence for causal associations between blood inflammatory proteins and the risk of Alzheimer's disease. In Chapter 7, I perform a systematic literature review to identify known blood protein correlates of Alzheimer's disease. I then use BayesR+ to conduct an integrated GWAS and EWAS on plasma levels of 282 Alzheimer's disease-associated proteins (n ≤ 1,064, GS). I observed strong evidence for causal associations between two proteins, TBCA and TREM2, and Alzheimer's disease risk. In Chapter 8, I examine associations between DNAm GrimAge and measures of brain health (n ≤ 709, LBC1936). A higher-than-expected DNAm GrimAge associated with poorer performance on cognitive tasks and neurostructural correlates of dementia at age 73. I observed weak evidence to suggest that DNAm GrimAge assessed at age 70 predicts cognitive decline up to age 79. In Chapter 9, I assess whether DNAm GrimAge and other measures of epigenetic ageing predict the prevalence and incidence of common disease states, including Alzheimer's disease (n ≤ 9,537, GS). Epigenetic ageing measures did not predict the prevalence or incidence of Alzheimer's disease. In Chapter 10, I discuss the major findings from this thesis in light of their limitations. The work presented in this thesis helps to detail the molecular regulation of 422 plasma protein levels and their causal associations with Alzheimer's disease. This work also highlights the performance of DNAm GrimAge in predicting indices of cognitive performance and common disease states. By incorporating genetic, epigenetic and protein data in two large-scale epidemiological studies, my findings inform our understanding of relationships between blood proteins and cognitive ageing and dementia.

The abrogation and de-regulation of cellular decisions in adult haematopoietic stem cells (HSCs) have been widely recognised as key contributing factors in ageing and disease. In particular, acquired mutations altering HSC self-renewal contribute to the formation of pre-leukaemic disorders, such as myeloproliferative neoplasms (MPNs), and ultimately to their transformation to acute myeloid leukaemia. Therefore, we urgently require a complete characterisation of the underlying molecular pathways of HSC self-renewal in order to provide crucial insights into disease progression and to help inform novel strategies for ex vivo expansion for gene therapy applications. To date, advances in functional and molecular single cell technologies have provided unprecedented resolution of heterogeneous HSC populations and their transcriptional landscapes. In contrast, the scarcity of functional HSCs and technical limitations of unbiased proteomic screening technology have prevented comprehensive characterisation of protein networks governing HSC fate. To overcome these technical limitations, we developed an optimised mass spectrometry workflow to interrogate as few as 10,000 HSCs and multiplex up to 16 cell fractions. Using this approach, we were able to quantify in excess of 4,000 proteins, while reducing the required cell input 30-fold. To identify key molecular drivers of in vivo HSC self-renewal, we probed HSC populations with increasing self-renewal potency, including TET2-deficient HSCs exhibiting a self-renewal advantage in MPN mouse models. Here, we observed reshaping of extracellular matrix protein networks, indicating a potential physical role of the neighbouring bone marrow niche cells for regulating HSC fate. Next, we integrated proteomic and transcriptomic to characterise the molecular pathways underlying HSC self-renewal across numerous -omics technologies. Here, we identified a wide range of intrinsic regulatory pathways and described molecular mechanisms regulating intracellular calcium levels in HSCs. In the final results chapter, we explored the ex vivo expansion of HSCs, since this is of paramount importance for the delivery of gene and cellular therapies against a plethora of monogenic haematological diseases. Recently pioneered HSC expansion protocols greatly enhanced the yield of phenotypic HSCs, but such cultures exhibit significant clone-to-clone variability in long-term self-renewal potency and differentiation. By linking the transplantation outcomes with transcriptional profiles of individual single cell-derived clones, we derived a novel gene signature for ex vivo expanded HSCs with long-term self-renewal potency and characterised key molecular pathways governing HSC fate throughout expansion. Furthermore, we identified a reporter strategy for prospectively isolating expanded HSCs. Together, these findings provided an insight into the molecular machinery underlying HSC self-renewal within the native bone marrow niche and during ex vivo expansion. Comprehensive multi-omic profiling also revealed the intricate relationship of gene expression profiles with the proteomic phenotype within the HSC compartment. Finally, we propose an optimised workflow for performing comprehensive proteomics on any rare cell populations which will be of use to researchers investigating a wide-range of cellular biology questions.

Human cytomegalovirus (HCMV) is a ubiquitous herpesvirus that represents a significant global health burden. In immunocompetent individuals, HCMV establishes a lifelong persistent and typically asymptomatic infection that is controlled by a multifaceted host immune response. However, immunocompromised patients, including transplant recipients and those with acquired immunodeficiency syndrome, are at a high risk of HCMV-associated morbidity and mortality. HCMV is also the most prevalent infectious cause of congenital disease, with the ability to cause neurodevelopmental complications. Cell-intrinsic immune responses are the first line of defence against viruses, mediated by constitutively expressed host proteins and processes that respond directly to virus infection. As a result of virus-host coevolution, viruses often antagonise antiviral host proteins by driving their downregulation, mislocalisation or inactivation. A quantitative proteomic analysis of HCMV infection published by our group found that at least 133 proteins are likely targeted for degradation by HCMV during early infection. For the project presented in this thesis, seven of these candidate antiviral factors were screened for antiviral activity using a novel restriction assay system and plaque assays. Two proteins, mixed lineage kinase domain-like pseudokinase (MLKL) and DmX-like protein 1 (DMXL1), were then selected for further mechanistic and functional characterisation. MLKL is the terminal effector of a form of cell death called necroptosis. Many herpesviruses suppress necroptotic signalling to evade cell death. However, the mechanism of HCMV-mediated inhibition of necroptosis has so far remained elusive. HCMV protein pUL36 was found to be necessary and sufficient for the downregulation of MLKL and the inhibition of necroptosis. pUL36 has previously been shown to inhibit another mode of cell death, apoptosis, making it a multifunctional cell death inhibitor. DMXL1 interacts with the vacuolar-type H+-ATPase to regulate endosomal acidification, with implications for endosomal trafficking, autophagy, immune signalling and many other cellular processes. The viral gene responsible for downregulation of DMXL1 was identified as pUS33A, which may recruit the E3 ligase Kip1 ubiquitination-promoting complex (KPC) to target DMXL1 for proteasomal degradation. Therapeutics currently available for treating HCMV are associated with significant toxicity and drug resistance. Characterisation of the protein-protein interactions underlying the viral evasion of cell-intrinsic immune responses may permit the development of small molecule therapeutics that disrupt these interactions and facilitate endogenous inhibition of viral replication. As well as contributing to our understanding of how HCMV regulates cell death and endosomal acidification, this thesis also presents proteomic data that will enable the identification of additional antiviral host proteins and the characterisation of virus protein function.

Centrosomes are small cytoplasmic organelles that play a critical role in multiple cellular processes; from mitotic spindle assembly and cell division, to cell migration, invasion and polarity. The human centrosome consists of two cylindrical structures known as centrioles, decorated at their distal ends by appendage structures, and surrounded by a proteinaceous matrix known as the pericentriolar material. Despite a number of advances in the field, focussing particularly on centrosome structure and assembly in normal and diseased cells, relatively little is known about the spatial and temporal dynamics of centrosome composition, and how this may vary between different cell, tissue, or disease types. Indeed, only a handful of studies have analysed the proteomic composition of the centrosome. In the first part of this study, I established a novel method for centrosome isolation that was developed in our laboratory. This approach, known as COMPACT (Centrosome Purification by Affinity Capture), involves a single-step affinity purification of centrosomes by a short peptide corresponding to a 33 amino acid C-terminal fragment of the centrosomal protein CCDC61. I have shown that the method is able to isolate centrosomes from a variety of cell types, and that analysis of purified centrosomes using mass spectrometry recovers the majority of known core centrosomal proteins. Importantly, I demonstrated that COMPACT purifies centrosomal components with higher efficiency and specificity than the traditional sucrose sedimentation-based centrosome isolation technique. In the second part of this study, I performed experiments to gain mechanistic insight into how COMPACT works. I generated cell lines lacking the centrosomal proteins, NIN or CEP128, and showed these proteins (located primarily at the subdistal appendages) to be directly involved in the peptide's ability to purify centrosomes. In the third part of this study, I performed a comprehensive analysis of the centrosome proteome in a panel of human cell lines, allowing for the identification of tissue-specific as well as differentially expressed centrosomal candidates. In addition to this, a quantitative mass spectrometry-based analysis of centrosome composition in different cell types revealed the absence of distal appendages in the T lymphocyte cell line, Jurkat, a novel and exciting finding. COMPACT was also shown to be effective at isolating centrosomes from primary tissue, revealing its potential for use as a tool to explore centrosome composition in various centrosome-associated diseases, including microcephaly and cancer. Finally, by comparing COMPACT across a wide range of cell lines, I was able to identify a number of novel centrosomal candidates, which were independently validated using various techniques. Loss-of-function studies in HEK 293T cells were performed to determine the function of a specific candidate, TRIM27. This study not only provides a novel method to isolate centrosomes in a spatially and temporally controlled manner, but also an effective tool for the identification of new or previously uncharacterised centrosomal proteins that may have significant roles in centrosome biology.

Next-generation sequencing technology has propelled the field of biology into the big data era, and continual advancements in computing have now made it easier to explore complex biological systems. However, analysing such highly complex data with conventional machine learning algorithms can be troublesome as these techniques require a considerable amount of feature engineering. Fortunately, a subfield of machine learning known as deep learning has recently, shown evidence towards overcoming these issues. Such algorithms have initially been applied in a genomic and transcriptomic setting. However, advancements in sequencing technology have allowed proteomics to mature to the level where deep learning is now a viable option. This thesis will primarily consider the application of deep learning for modelling the various properties of a protein. Although deep learning addresses some of the initial problems encountered when analysing omic data, there is a series of different challenges that are still present when applying deep learning algorithms. Even when using the latest approaches, a deep learning model often requires a large amount of labelled data, which can be costly and time-consuming to acquire. If there is not an adequate amount of data, then standard deep learning approaches often underperform when compared to traditional machine learning algorithms. Also, these models are black-box algorithms, which presents problems with the interpretation of the predictions being produced by the model. Given the variability within proteins, it can be difficult to summarise the data effectively as information about the protein can be lost through feature engineering. In each research chapter of this thesis, we address the shortcomings of applying conventional machine learning to model protein data using deep learning. In the first technical chapter, we begin with the use of state-of-art subword encoding schemes. We prove that these new representations are more beneficial and practical for pre-training when compared to the standard baselines. In the next chapter, we go a step further and address the issue of applying deep learning models to smaller datasets. In doing so, we explore how metric learning can be used to form a robust model architecture that is capable of learning and ranking proteins from just a few labelled examples. After this, we consider an approach that utilises both pre-training and metric learning to reach a new state-of-the-art by using large unsupervised networks. In this chapter, we leverage a BERT model that has been pre-trained on a vast quantity of proteomic data, to model a collection of regression tasks using only a minimal amount of data. We adopt a triplet network structure to fine-tune the BERT model for each dataset and evaluate its performance on a set of downstream tasks. The first three strategies mentioned were tested on a variety of downstream tasks: four protein property prediction tasks (plasma membrane localisation, thermostability, peak absorption wavelength, enantioselectivity). Additionally, this thesis includes a further two chapters that consider other challenges encountered when modelling protein data. This begins with a chapter on applying pre-training to improve upon the state-of-the-art in phosphorylation site modelling using a brand-new convolutional transformer-based model. We evaluate our approach on a general phosphorylation site dataset, and a variety of kinase-specific datasets. Additionally, to emphasise that this is an example of white-box deep learning, we visualise the model's features to gain a better understanding behind the prediction of each site. The final research chapter considers a state-of-the-art approach to modelling the interactions between proteins and drugs. In this chapter, we leverage a set of BERT-style models that have been pre-trained on vast quantities of both protein and drug data. The encodings produced by each model are then utilised as node representations for a graph convolutional neural network, which in turn models the interactions without the need to simultaneously fine-tune both protein and drug BERT models to the task. We evaluate the performance of our approach on two drug-target interaction datasets that were previously used as benchmarks in recent work.

Cells respond to stress by up-regulating chaperone proteins that correct protein misfolding to maintain function. However, protein homeostasis is lost in ageing, leading to aggregates characteristic of protein-folding diseases. While much is known about how these diseases progress, discovering what causes protein-folding to deteriorate could be key to their prevention. Studying ageing in humans introduces a key temporal obstacle, however results in model organisms suggest ageing has cellular origins. In this work, primary human mesenchymal stem cells (hMSCs) have been cultured to a point of replicative senescence and subjected to proteotoxic stress, as an in vitro model to investigate changes to the stress response during cellular ageing. Through -omic analyses it has been shown that the maintenance of protein homeostasis deteriorated in senescent cells through the attenuation of processes downstream of chaperone transcription. This was coincident with lowered levels of a functional module of chaperone proteins associated with stress-inducible heat shock protein 70 kDa (HSPA1A). Further analysis of the temporal dynamics of the transcriptomic and proteomic stress responses revealed a lack of translational capacity to be a limiting factor in the ability of senescent cells to mitigate proteotoxic stress, leading to a loss of speed, magnitude, and efficacy of the cellular stress response. Furthermore, here it has been shown that senescent cells experience partial loss of the E3 ubiquitin ligase CHIP. CHIP is known to regulate HSPA1A stability during stress and cooperates with the chaperone to mark misfolded proteins for turnover. By building a mathematical model of the proteotoxic stress response, in silico simulations have predicted that a decrease in cellular CHIP leads to a greater accumulation of misfolded proteins during proteotoxic stress. Using a cysteine-reactive label which binds to residues typically buried in correctly folded proteins, this prediction has been validated by showing significant conformational compromise to the proteomes of senescent cells in response to stress, whereas no perturbation was seen in proliferating cell populations. Prime among the conformationally compromised proteins in senescent cells were several cytoskeletal proteins. Here it has been shown that this is likely due to the senescence-associated down-regulation of another functional module of chaperone proteins responsible for maintaining cytoskeletal protein homeostasis, the chaperonin-containing tailless (CCT) complex. These results demonstrate multiple mechanisms which drive the attenuation of the proteotoxic stress response in cellular ageing. As such, these represent a step forward in our understanding of the underlying link between ageing and a loss of protein homeostasis. Given the therapeutic potential of hMSCs, this work also holds relevance to the emerging field of regenerative medicine. In this context, this study gives a comprehensive evaluation of consequential, but often neglected, changes to the hMSC stress response following extended culture in vitro.

Type 2 diabetes (T2D) is a complex chronic disease which is associated with micro- and macro-vascular complications and premature mortality. This multifactorial disease originates from a complex interplay between genetic, behavioural and environmental risk factors, the underlying mechanisms of which are unclear. The goal of this thesis was to investigate the contributions of genetic, metabolic and behavioural risk factors in the aetiology of T2D by studying large-scale population studies, in order to provide insights for the development of T2D prevention and intervention strategies. I first conducted a genome-wide association study on T2D in a large prospective cohort study for T2D, the EPIC-InterAct study. This work provides a valuable resource for the research community to conduct in-depth genetic investigations on T2D. The strongest T2D-associated genetic variant was at the established TCF7L2 locus. I then used the profiling of behavioural factors in the EPIC-InterAct study to show that there was no interaction between TCF7L2 genetic variants and an index of the number of health behaviour goals achieved, a measure that was strongly associated with diabetes risk. In a meta-analysis of three randomised controlled trials designed to investigate the effect of lifestyle intervention on the risk of developing T2D, I showed that there was no overall interaction between TCF7L2 and allocation to the intervention rather than the control arm. However, there was an interaction in one trial, which focused more on physical activity and dietary change than on weight change. In an analysis of a detailed quantitative metabolic trait study, I demonstrated a specific interaction between TCF7L2 and physical activity energy expenditure (PAEE). These data suggest that genetic targeting could have additive effects with health behaviour interventions and may be more valuable for personalised prevention if linked to very specific behaviours. To build on this observation and explore the potential mechanisms through which physical activity may affect the risk of T2D in different genetic subgroups, I undertook a systematic investigation of interaction effects between PAEE and genetic risk scores for glycaemic traits in the Fenland study. I did not find any significant interaction. Alternatively, using an agnostic approach, I searched for potential interactions with PAEE across the genome. Preliminary analyses detected three variants showing genome-wide significant interaction effects, although the mechanisms were unclear. As obesity is a strong risk factor for T2D and a previous study had reported that BMI modifies the effect of LAMA1 gene on T2D risk, I undertook an in-depth analysis of this potential interaction in the large-scale UKBB and EPIC-InterAct studies. I showed and confirmed that LAMA1 has a stronger effect on T2D among lean individuals compared with obese individuals in independent studies. Observational studies suggest that low cardiovascular fitness (CRF) is an independent T2D risk factor. However, it is unclear whether the association is causal or is due to confounding. I performed the largest genetic discovery of cardiorespiratory fitness in the UKBB study and developed a genetic risk score for CRF. In Mendelian randomisation analyses, I showed that the associations of this score with T2D and hyperinsulinaemia were compatible with underlying causal relationships. High-throughput metabolomics based on large-scale population studies offers a great opportunity for providing direct biological insights and identifying physiological alterations linking risk factors to T2D. I characterised the metabolic profiles of physical activity and cardiorespiratory fitness and identified shared and distinct signatures highlighting the roles of branched-chain amino acid metabolism and fatty acid oxidation. Overall, my PhD work contributes towards 1) a wider understanding of interpersonal differences for diabetes risk in response to behavioural factors based on variable genetic make-up and its implication in population-level diabetes prevention and intervention strategies 2) the discovery of the genetic basis of cardiorespiratory fitness and the provision of strong evidence supporting its causal role for T2D; 3) the characterisation of metabolic profiles of physical activity and cardiorespiratory fitness and the generation of hypotheses for potential biological pathways and physiological mechanisms underlying their influences on diabetes development.

The endosomal network is a crucial sorting hub in eukaryotic cells, responsible for integrating transmembrane proteins and lipids, termed 'cargoes', from multiple input pathways and sorting them for a crucial subsequent fate decision. Cargoes in the endosomal network can either be degraded within the lysosome, or recycled back to an acceptor compartment such as the plasma membrane or trans-Golgi network (TGN). Endosomal sorting complexes that facilitate this process play as central role in the maintenance of cellular homeostasis by regulating a delicate balance between cargo degradation and recycling. Perturbations to endosomal recycling are increasingly associated with disease, many of which are neurodegenerative in their aetiology. Recent methodological advances have expanded the understanding of the flux of cargoes through the endosomal network, with hundreds of transmembrane proteins depending upon sequence-dependent sorting for their delivery to the plasma membrane. In this thesis, I utilised proteomics and RNA-sequencing techniques to investigate two key complexes involved in orchestrating endosomal recycling: the retromer complex, and the endosomal SNX-BAR sorting complex promoting exit-1 (ESCPE-1). I identified a multivariate phenotype of endolysosomal dysfunction upon retromer depletion, consistent with complex neurodegenerative phenotypes reported in the literature. By developing a methodology to specifically label trans¬-Golgi network proteins by proximity biotinylation, I also performed a screen to identify retrograde endosome-to-TGN cargo proteins for ESCPE-1, which highlighted Neuropilin-1 as a novel cargo. The coronavirus disease 2019 (COVID-19) pandemic emerged in 2019 and swept across the globe, causing > 116,000,000 cases and > 2,500,000 deaths worldwide at the time of writing. As part of a collaborative project, we identified Neuropilin-1 as an important host factor for infection by the causative virus, SARS-CoV-2. We showed that SARS-CoV-2 directly binds to Neuropilin-1, and blocking this interaction suppresses infection in cell culture, therefore establishing Neuropilin-1 as an important therapeutic target in the study of COVID-19.

Introduction: Cardiovascular diseases remain a major cause of morbidity and mortality worldwide. There is a large number of patients who suffer from myocardial infarction (MI). However, the a hearts of majority of patients who survive an MI will undergo substantial changes that will lead to heart failure (HF). The aim of this work is to characterise post ischaemic cardiac remodelling in two different experimental models using rats and pigs. A clinical study was also planned to identify acute changes in the blood metabolome of patients undergoing primary percutaneous coronary intervention (PPCI) following ST elevation myocardial infarction (STEMI) and correlate with long term recovery. The obtained information will help in identifying targets for therapeutic interventions, which can reduce the likelihood of adverse cardiac remodelling, and therefore heart failure. Methods: Adult rats undergoing left anterior descending (LAD) coronary artery ligation and pigs with LAD coronary artery balloon occlusion were used in this work. SHAM operated rats and healthy pigs were used as control. After initiation of cardiac injury, the animals were monitored for 4 weeks before termination. Cardiac injury and function 4 weeks after interventions were assessed using echocardiography and cardiac proteins release in pigs. Hearts and blood samples were extracted and collected. Nuclear magnetic resonance spectroscopy was used to assess metabolic profile of blood and cardiac tissues. Cardiac energy metabolites were measured using high performance liquid chromatography. Cardiac protein and phospho-protein quantification were carried out using liquid chromatography tandem mass spectrometry. Histological analyses of infarcted and control hearts were conducted using light and electron microscopies to detect ultrastructural differences. Ethical application and protocol for the clinical study were granted. Results: Post ischaemic cardiac remodelling in rodent and porcine hearts was associated with increased cardiac fibrosis and wall thickening of small to medium size arterioles. Ultrastructural analysis showed characteristics of ongoing structural remodelling in the extracellular matrix and the presence of fibroblasts and telocytes. There was disruption to mitochondrial cristae density and distribution. No significant changes were detected in cardiac energetics in infarcted and control tissues of rats and pigs. Proteomics and phosphoproteomics analyses revealed evidence of sustained oxidative stress and inflammation, in addition decrease of mitochondrial respiration and metabolism proteins, modification to calcium cycling, z disc and selected extracellular matrix proteins and collagen subtypes. There was increase in certain metabolites with post ischaemic remodelling, mainly ketone bodies in rodents and porcine blood, and taurine in porcine cardiac tissues. The clinical study was suspended due to Covid-19 pandemic. Conclusion: Post ischaemic cardiac remodelling is associated with substantial cellular, molecular and metabolic changes in rodent and porcine models of cardiac injury. Importantly, the increase in blood ketone bodies in both models, in addition to decreased porcine blood taurine level could potentially be investigated furthermore in patients to target adverse cardiac remodelling and design therapeutic intervention for heart failure.

Cardiovascular disease is the number one cause of death globally, leading to high proportion of premature deaths. Since there is an aging population, the economic burden on health care systems increasingly requires the discovery of new biomarkers. Plasma proteomics is an approach to identify novel biomarkers using plasma. However, there are limitations with plasma due to the complexity and dynamic range between high abundant and low abundant proteins. Approaches of overcoming complexity and dynamic range include Molecular imprinting which is a promising tool in the field of proteomics. This thesis has created an imprint using polystyrene as substrate, which can be used to enrich albumin specifically with polymerised dopamine offering huge advantages such as polymerisation in the presence of urea. Dopamine does not require complex chemicals, scalable, cheap and biodegradable. Mouse heart organ was also used to create imprints to heart proteins onto polystyrene coated with polydopamine. Mass spectrometry was used to analyse proteins enriched from human plasma. A Nano-LC system was coupled to an Orbitrap (Q-Exactive) providing both quantitative and qualitative data. Both shotgun and targeted proteomics were employed. The Q-Exactive was operated using parallel reaction monitoring which utilises the high resolution capability of the Q-Exactive to quantitatively target peptides of proteins. Shotgun proteomics using data dependent acquisition was also employed to assess all proteins involved in the imprinting and enrichment. Extracts from molecular imprints (MIPs) contacted with plasma and their respective non-imprint controls (NIPs) were analysed using mass spectrometry. Results from mouse organ MIP contacted with plasma from five myocardial infarction patients, showed the presence of: Troponin, Fatty acid binding protein 3, Creatine Kinase, Lactate Dehydrogenase, Cardiac myosin binding protein C. The imprint created from tissue and organ shows the potential to analyse other tissue and organ MIPs to specifically study organ damage in other diseases, for example kidney, liver and lung injury related to COVID-19 and the effects of COVID-19 on organs.

The mechanism for human labour remains poorly understood, restricted by difficulties in human pregnancy experimentation, and this limits our ability to manage complications of parturition including spontaneous preterm birth and induction of labour. The studies presented in this thesis utilise metabolomic and phosphoproteomic techniques to improve our understanding of the pathways involved in the trigger for labour in humans. The first study in the thesis compared mass spectrometry metabolomics analysis of maternal and cord plasma at the time of delivery, and the results indicate that the trigger for spontaneous labour in humans may involve endocannabinoid, sphingolipid and ceramide pathways. The second study presented investigates this further, comparing the metabolic profiles of maternal, cord artery, and cord vein blood sampled from women during low-risk pregnancies, including comparisons of metabolite concentrations between different labour onsets and delivery. This study indicates that there are inherent differences in fetal and maternal signals between women who spontaneously labour and women who have an induction of labour. The third study assessed potential associations between the maternal metabolic profile measured in the second trimester with gestational age at delivery within the Born in Bradford cohort study (analysis sample n = 7440). This study indicates a metabolic profile of dyslipidaemia may be associated with shorter gestational age at delivery, and that non-pathological human labour may involve pathways related to inflammation and vascular function. The final study used phosphoproteomics analysis of fresh myometrium in vitro, and the findings indicate that there may be a myometrial phenotype associated with failed induction of labour and failure to progress. Taken together, these results provide novel leads to follow in order to progress our quest for a full understanding of the trigger for spontaneous labour in humans.

Proteins are biomolecules that govern the biochemical processes of the cell. Correct cellular function, therefore, depends on correct protein function. For a protein to function as intended, there need to be sufficient copies of that protein, it should be correctly folded into its tertiary structure and ought to be in proximity of its interaction partners, amongst many other requirements. For a protein to be in the proximity of its interaction partners, whether those be other proteins, RNA or metabolites, it needs to be localised to the required compartment. Cells from all organisms display sub-cellular compartmentalisation, though to vastly differing degrees. E.Coli, for example, has remarkably simple sub-cellular organisation, whilst the apicomplexan Toxoplasma gondii has a vast number of specialised organelles. In seminal experiments, Christian De Duve showed that upon biochemical fractionation of the cell, proteins co-fractionated if they were localised to the same organelle. These experiments led to the discovery of two organelles: the lysosome and the peroxisome, for which Christian De Duve was awarded the Nobel prize. Upon the advent of mass-spectrometry, these experiments were refashioned into high-throughput techniques with the development of Localisation of Organelle Proteins by Isotope Tagging (LOPIT) and Protein Correlation Profiling (PCP). Now these techniques have been redeveloped and a typical experiment can accurately measure thousands of proteins per experiments, whilst also providing information on (at least) a dozen subcellular compartments. To analyse spatial proteomics data, they are first annotated with marker proteins, which are proteins with a priori known unambiguous localisations. Typical analysis proceeds by training a machine learning classifier to assigned proteins with unknown localisations to one of the compartments based on the spatial proteomics data. However, this framework holds back spatial proteomics from answering more complex questions. The first challenge is that proteins are not necessarily localised to a single compartment and so there is uncertainty associating a protein with an organelle. There is also uncertainty associated with the experiment itself, for example, reproducing the biochemical fraction and the stochastic nature of mass spectrometric quantitation. Two chapters of my thesis are dedicated to alleviating this problem by developing a Bayesian model for spatial proteomics data, with dedicated software. These approaches perform competitively with state-of-the-art classification algorithms whilst Markov-chain Monte Carlo algorithms are employed to sample from the posterior distribution of localisation probabilities. This is the basis for quantifying uncertainty in protein-organelle associations. This Bayesian approach has several limitations, for example it still relies on marker proteins. This precludes analysis of poorly annotated non-model organisms using spatial proteomics techniques. A chapter of my thesis is dedicated to this challenge with a motivating application to the T. gondii sub-proteome. Following on from this in a separate chapter, I develop a semi-supervised Bayesian model that reduces the reliance on marker proteins. The application to T. gondii constitutes a massive knowledge expansion revealing localisation of thousands of proteins to complex specialised niches. I also analyse the relative redundancy of the organelle sub-proteomes and the selective pressure of the host-adaptive response, revealing previously unknown insights. The semi-supervised Bayesian approach makes use of the principle of over-fitted mixtures, currently used for data clustering, by extending it to model spatial proteomics data. Reanalysis of spatial proteomics data reveals new annotations in all datasets and allows interrogation of previously overlooked organelles. Another limitation of the approaches, thus far, is the parametric assumptions made by the Bayesian approach. One chapter is dedicated to placing the analysis of spatial proteomics in the semi-supervised Bayesian non-parametric context. In the final chapters of thesis, I summarise the modern questions that spatial proteomics seeks to answer, including deciphering multi-localisation, change in localisation and the effect of post-translation modifications on subcellular localisation. I carefully define these problems and motivate further Bayesian models. I develop a Bayesian model to analyse differential localisation experiments; that is, spatial proteomics concerned with changes in localisation. This approach improves over currently ad-hoc methods applied to such data. I conclude with the limitations of our approach and potential solutions to the other methods.

Protein glycosylation is present in all kingdoms of life, from small virus particles to complex living organisms. It is the most abundant and diverse form of post-translational modification, ranging from the attachment of a single monosaccharide to entire sugar chains onto cellular proteins. This modification affects the recognition and interaction of proteins with their interaction partners, altering their role in health and diseases. Despite its importance, the study of protein glycosylation has lacked behind other biomolecules and post-translational modifications, partly because of absence of available tools and methods. This doctoral thesis aims at expanding existing analytical approaches and developing new technologies to study the different types of protein glycosylation using tandem mass spectrometry. The second chapter demonstrates de novo determination of the sequence and protein N-glycosylation of the anti- H2AX monoclonal antibody, enabling structural modelling of the antigen recognition and laying the groundwork for further development of this diagnostic reagent for DNA damage in cancers. Chapter 3 focusses on improving analysis of protein O-glycosylation using tandem mass spectrometry. The use of low-energy stepped collision energy (SCE) methods is described for improved characterisation of the glycan part and site-localisation of protein O-glycosylation. SCE-HCD methods such as 10-20% of normalised collision energy (NCE) show the formation of valuable fragment ions enabling a more accurate identification of O-glycopeptides. The O-GlcNAc modification is a particular form of O-glycosylation, consisting in the attachment of a single N-acetylglucosamine sugar to proteins. In chapter 4, the role of O-GlcNAc as an epigenetic mark on H2B-S36 is investigated. Homogeneously GlcNAcylated histones and nucleosomes were synthesised and used in pull-down ex- periments to enrich putative interaction partners for the H2B-S36-GlcNAc modifica- tion. Mass spectrometry-based quantitative interaction proteomics revealed a direct interaction between H2B-S36-O-GlcNAcylated nucleosomes and the TCF20 transcrip- tion factor, suggesting a possible signalling role for the O-GlcNAc modification. Lastly, chapter 5 shows that synthetic histones with reactive side chains can be used to co- valently trap interaction partners in a native context, identifying their binding site by crosslinking mass spectrometry with amino-acid resolution inside the interaction.

Breast cancer is the most common cancer in women in the western world. It has been shown that estrogen receptor alpha (ERα) contributes to tumour formation in 75% of breast cancer cases and various drugs, such as 4-hydroxytamoxifen (OHT), have been used to inhibit estrogen-induced signals. Studies have highlighted the importance of interactions between ERα and various factors in regulating chromatin binding and gene transcription, underlying estrogen signalling in breast cancer. During the last decade, quantitative mass spectrometry (MS)-based proteomics has become a powerful tool for the study of protein interactions. However, technical challenges have restricted most applications to qualitative observations and the dynamics of protein interactomes have until recently been largely unexplored. Here, we describe the development and application of a quantitative multiplexed method, termed qPLEX-RIME, which integrates RIME with isobaric labelling for the study of protein interactome dynamics in a quantitative fashion with increased sensitivity. Using the qPLEX-RIME method we successfully identified endogenous ERα-associated proteins in human Patient Derived Xenograft (PDX) and primary tumours and we delineated the mechanistic temporal changes of the ERα interactome in MCF7 cells upon OHT treatment. Our pipeline was also successfully applied for the study of other factors, such as CBP, NCOA3 and RNA Polymerase II. The integration of the different interactome datasets, lead to the discovery of a transcription factor, termed ZNF207, that belongs to the zinc finger protein family. In MCF7 cells, ZNF207 interacts with ERα, known ERα interactors and components of the transcription machinery. Knockdown of ZNF207, affected cellular proliferation of various cancer and non-cancer cell models, suggesting a functional role in the majority of cellular contexts. Interactome and chromatin binding assays revealed that ZNF207 is important for the assembly and chromatin binding of the Mediator complex and general transcription factors, whilst RNA-sequencing and proteomic analysis showed a decrease in expression of genes linked to cell cycle followed ZNF207 knockdown. The effect was likely mediated by changes in the chromatin binding of the subunit MED1 and the transcription factor TAF3 to promoters of cell cycle genes. Thus, ZNF207 plays an important role in our model system and was studied in depth in this thesis by integrating various functional assays. Our results demonstrate that qPLEX-RIME offers a powerful tool for the in-depth characterisation of protein interactome dynamics, which is also applicable to clinical samples. Our findings constitute the basis for a deeper mechanistic insight into the dynamics of ERα interaction network and its role in breast cancer along with the discovery of a transcription factor that may play a vital role in the initiation of transcription.

Through a combination of epidemiological and proteomic methods, the aim of this thesis was to advance understanding of antibody-driven immune responses that partially protect against infection and reduce fecundity in human schistosome infections. The relative contribution of exposure to infection and IgE antibody responses against schistosomes were explored using a mathematical modelling approach. Age intensity and serology profiles of sub-cohorts were reproduced to good approximation by incorporating explicitly defined exposure and immunity functions, fitted to field data, into an age- and sex-structured S. mansoni transmission model. However, the inferior model fit for males from one tribal group highlights the need to capture heterogeneity in antibody responses that occur with varying cumulative exposure of different demographic groups. The progressive development of protective immunity against S. mansoni infection is associated with cross-reactivity between members of the Tegument Allergen-Like (TAL) protein family. This relationship was explored further and evidence for involvement of an additional TAL family member in the development of IgE-mediated anti-infection immunity found. Evidence supporting similar cross-reactive relationships between S. haematobium TAL proteins is provided. In addition to anti-infection immunity, S. haematobium is thought subject to an anti-fecundity immune response. It is proposed that IgG1 antibodies raised to proteins within whole worm extract are responsible. A proteomic approach used to identify targets of the IgG1 returned 191 proteins, several of which are closely associated with the parasite-host interface. Three were successfully expressed and analysis identified possible associations between target specific IgG1 and reduced worm fecundity, although this was strongly confounded by host age and village of residence, a proxy for the force of transmission. These findings all demonstrate associations between the development of protective antibody responses and variations in exposure behaviour, host age or the force of transmission; the combined contribution of which will determine the level of exposure to immunostimulatory antigens and ultimately development of immunity.

Pseudomonas aeruginosa (Pa) is a predominant pathogen in the airways of people with cystic fibrosis (CF). Pa undergoes several evolutionary adaptations that allow it to thrive in the CF airway environment, including the emergence of amino acid auxotrophs. Arginine auxotrophs arise at high frequency in CF, and arginine is known to play an essential physiological role in Pa, so I chose to investigate the possible molecular basis of this further. I found that a transposon (Tn) mutant containing a disrupted argA gene was auxotrophic for arginine, but readily accrued secondary mutations enabling the mutant to grow, even in the absence of this amino acid. Whole genome sequencing revealed that one of these bypass mutants (denoted hereafter, AMB128) contained a single nucleotide polymorphism (SNP) in the ambA gene. Complementation analysis confirmed that the SNP in ambA was responsible for the arginine auxotrophy bypass, and further investigation established that the SNP conferred change-of-function rather than loss-of-function to the AmbA protein. Detailed proteomic profiling of AMB128 revealed up-regulation of certain peptides derived from the enzyme, N-acetyl glutamate synthase (NAGS) encoded by argA. This was unexpected because the argA gene in AMB128 was disrupted by a Tn insertion (and this insertion was confirmed by inspection of the whole genome sequence). This up-regulation was found to be likely mediated by an outward-facing promoter in the Tn sequence, driving expression of a truncated regulatory domain and functional catalytic GNAT domain in the NAGS. The truncated regulatory domain was anticipated to disrupt feedback inhibition of the enzyme. Counter-intuitively, the Tn insertion may therefore elevate the level(s) of certain intermediates in this arginine biosynthetic pathway. The proteomic analysis also provided insight into the likely cellular response to the accumulation of these metabolic intermediates. Proteins involved in polyamine catabolism and transport of small molecules were up-regulated in AMB128, whereas many proteins involved in denitrification were down-regulated. In summary, this work demonstrates remarkable metabolic plasticity of Pa by highlighting one mechanism by which the organism can adapt when faced with dysregulated arginine biosynthesis. In this regard, it is worth noting that the ambA gene in a selection of other spontaneous argA::Tn bypass mutants did not contain any SNPs, indicating that this is just one of (likely) many possible evolutionary “solutions” to the problem.

With the improvement of high-throughput technologies during the last decade, several studies exploring the normal gene expression in human tissues have been published. Many studies examine the transcriptome with RNA sequencing (RNA-Seq), and others probe the proteome with unlabelled bottom-up Mass Spectrometry. As the sampling of undiseased tissues is difficult, the community often refers to expression atlases, which are collating these studies, to support or validate new findings. Despite many overlapping tissues between the studies, few atlases attempt to integrate all the data. In this thesis, I investigate the consistency of gene expression across tissues and studies in human with the help of transcriptomics captured with high-throughput sequencing (RNA-Seq) and proteomics generated with label-free bottom-up Mass Spectrometry (MS). After describing the transcriptomic and proteomic data and their state-of-art processing (Chapter 2), I review several identified sources of biases and my approaches to limit their effects (Chapter 3). The integration of the various transcriptomic datasets (Chapter 4) shows that the biological signal dominates the technical noise for RNA-Seq data. Tissue samples display higher levels of correlation for identical tissues in other studies than for other tissues in the same datasets. In other words, interstudy correlations for identical tissues are higher than correlations between different tissues within the same study. Globally, genes show similar expression profiles across studies for a given set of tissues. All genes categories are involved, including the tissue-specific genes and the ubiquitously expressed ones. After briefly discussing comparisons of proteomic data, I introduce a new proteomic quantification method, PPKM (Chapter 5). The PPKM method allows me to quantify about twice as many proteins compared to usual methods. Limited numbers of previous studies have shown various correlation levels between the expression of protein and mRNA in studies combining high-throughput transcriptomics and proteomics. I show that, for most tissues, we can observe quite good correlation levels (i.e. significantly better than expected by chance), even when the samples have different biological and technical backgrounds as they have been independently sourced. Many genes share similar patterns of expression between the two biological layers, e.g. genes that have a protein detected in a single tissue are more likely to have their mRNA showing specificity for the same tissue. Additionally, three groups of genes present functional enrichments of biological processes. Genes having highly correlated protein and mRNA expressions across tissues are enriched in catabolic processes. Genes having the most anticorrelated expressions are enriched for ribosomes and ncRNAs regulation. Genes with a protein detected in a single tissue are enriched in signalling processes. Overall, this thesis describes a global picture of the current consolidated knowledge we can extract from the joint study of public transcriptomic and proteomic data. Beyond confirming or improving observations reported in the literature, this work provides new insights into the ubiquitous and tissue-specific genes. To the best of my knowledge, this work has also established the most extensive list of genes with robust transcriptomic and proteomic expression across tissues and studies. Furthermore, it shows that joint study approaches can help the development of new methods, like the new proteomic PPKM quantification method. Finally, the highlighting of distinct functional enrichment profiles for groups of genes across tissues and studies lays a framework for further research.

Eukaryotic cells are characterised by a high degree of structural partitioning into distinct microenvironments dedicated to diverse and specific roles. To maintain these microenvironments, the functions of endogenous RNAs and proteins are fine-tuned by a set of. post-transcriptional and post-translational regulation mechanisms fundamental to complex cellular processes, throughout the stages of development or in response to acute stimuli. Whereas microscopy- or affinity-based methods are routinely employed to determine the location or the interactions of a limited number of proteins or RNAs, a system-wide analysis enables a thorough understanding of the functional and spatial dynamics of the proteome and transcriptome. In this Thesis, two biochemical fractionation-based techniques were developed in the quest for subcellular exploration of the proteome and transcriptome. In addition, two biological questions were investigated with regard to the post-transcriptional control of the ultraviolet irradiation induced stress response and stem cell pluripotency phase transition. In Chapter 3, a one-step subcellular fractionation procedure using multiple iodixanol containing gradients, termed "MultiLOPIT", is presented, which provides high-resolution organelle resolution for comparative applications in different cell states or conditions. In Chapter 4, characterisation of the dynamics of protein and RNA-binding proteins, post-translational modification, and spatiotemporal localisation are investigated in response to UV-induced cellular stress. This multi-omics approach highlights selective cytosolic translation programmes and chaperone machinery upon UV irradiation. Chapter 5 describes an efficient, high-throughput and scalable method for simultaneous mapping of the subcellular transcriptome, proteome and RBPome. Finally, chapter 6 delineates the dynamics of the post-transcriptional regulatory programmes mediated with RBPs, highlighting ribosome biogenesis and splicing mechanism involvement through the exit of naïve pluripotency. The results of the research in this Thesis offer a strong basis for the application of localisation, interactions and dynamics studies of RNA and proteins, which enable multi-layered system-level interrogations of cell biology.

Major depressive disorder (MDD) is a highly prevalent and disabling condition with a complex pathophysiology that has not been fully elucidated to date. While the socioeconomic burden of the disease is significant, many individuals remain undiagnosed or misdiagnosed. This is largely because the current diagnostic approach that relies on clinical evaluations of signs and symptoms can be subjective, and time and resources tend to be rather limited in primary care where the majority seek help for depression. Therefore, there is a significant and pressing need for an objective, reliable and readily accessible diagnostic test to enable earlier and more accurate diagnosis of MDD. In particular, as individuals experiencing subthreshold levels of depressive symptoms have an increased risk of developing MDD, it would be clinically relevant for such a diagnostic test to be able to identify depressed patients and/or individuals with high risks of incident MDD among symptomatic individuals. This thesis sought to develop risk prediction models that could potentially be utilised within a clinical setting to facilitate earlier and more accurate diagnosis of MDD. Such models were used to obtain probability estimates of the investigated individuals having or developing MDD based on their blood-based proteomic profiles and other characteristics, including sociodemographic and lifestyle factors. A targeted mass spectrometry approach was used to measure the abundances of a panel of peptides representing proteins, many of which have been previously associated with psychiatric disorders. Biomarkers were investigated in serum samples, which are widely used for blood-based biomarker discovery, as well as in dried blood spot samples, which are relatively novel in the field and carry several advantages. Importantly, this thesis focused on adopting appropriate statistical methods to ensure that the diagnostic predictions made by the models were accurate and reproducible, by addressing problems of model overfitting and model selection uncertainty. A particularly significant aspect of this was the development and application of a multimodel-based approach combining feature extraction and model averaging, which resulted in improved model predictive performance and generalisability. Diagnostic prediction models based on serum proteomic, sociodemographic/lifestyle and clinical data were shown to be able to differentiate between subthreshold symptomatic individuals who developed and did not develop MDD. Additionally, diagnostic prediction models based on dried blood spot proteomic and digital mental health assessment data were shown to be able to identify currently depressed patients without an existing MDD diagnosis as well as currently not depressed patients with an existing MDD diagnosis among subthreshold symptomatic individuals. These results clearly demonstrate the potential of such prediction models to be used as an aid to the diagnosis of MDD in clinical practice, especially within the primary care setting. Moreover, MDD was found to be associated with several blood-based proteomic biomarkers, which mainly represented an immune/inflammatory profile, as well as with various other patient features, most notably body mass index and childhood trauma. Although further investigations are needed, these associations reveal disturbances in the stress response pathways involving the hypothalamic-pituitary-adrenal axis in the pathophysiology of depression.

Pancreatic ductal adenocarcinoma (PDAC) has the lowest survival rate of all major cancers, and is characterised by an intense desmoplastic reaction. The upregulation and remodelling of extracellular matrix (ECM) by activated fibroblasts results in an extremely rigid tumour environment that contributes to tumorigenic processes such as cell proliferation and invasion. Cells mediate attachment to the ECM through integrin adhesion complexes (IACs). IACs sense external and internal mechanical cues, such as matrix rigidity and actomyosin contractility, to govern cell behaviour through a variety of IAC-associated molecules, collectively known as the adhesome. However, precisely how IACs regulate cell behaviour in response to force is not fully understood. To address this, a proximity-dependent labelling method, BioID, was used to study force- regulated changes in adhesome interaction networks in a disease-relevant cell line. A protocol was developed and optimised for the detection of proximity interactions in mouse pancreatic fibroblasts, and using 16 IAC bait proteins, label-free quantitative proteomics was performed to define a proximity-dependent adhesome under standard tissue culture conditions. Stringent analysis revealed a network of 147 enriched proteins with 360 proximity interactions. Many known IAC components were identified, together with candidates that may have underappreciated roles in adhesion. Unbiased analysis revealed 5 clusters of IAC BioID baits representing groups of proximal proteins that link to common groups of prey, which may represent distinct functional modules within IACs. To examine force-dependent interactions in the proximity-dependent adhesome, actomyosin contractility was disrupted through myosin II inhibition with blebbistatin, resulting in a loss of actin stress fibres and an enrichment of nascent adhesions. The effects of substrate rigidity on the proximity-dependent adhesome were also examined, with cells seeded on stiff substrates reflecting the cancerous pancreatic environment exhibiting increased proliferation and larger IACs compared to those grown on soft substrates that reflect the normal pancreas. A number of actomyosin-dependent and substrate-dependent interactions were identified using BioID, providing some insight into mechanoregulatory mechanisms within the adhesome. Alongside multiple established mechanoresponsive components, several potential new candidates were identified that may be involved in mechanotransduction by mouse pancreatic fibroblasts. Future experiments will probe these mechanisms further. This study presents the first empirically defined view of proximal interactions in the adhesome on a global level, and has demonstrated that BioID is a valuable tool that complements existing methods to study the effects of force on IACs. It is hoped that an increased understanding of how adhesome networks are regulated may provide insights into the relationship between the ECM and pancreatic cancer.