Search

Understanding the mechanisms underlying tumour heterogeneity is key to the development of treatments that can target specific tumour subtypes. We have previously targeted CRE recombinase-dependent conditional deletion of the tumour suppressor genes Brca1, Brca2, p53 (also known as Trp53) and/or Pten to basal or luminal oestrogen receptor-negative (ER−) cells of the mouse mammary epithelium. We demonstrated that both the cell-of-origin and the tumour-initiating genetic lesions cooperate to influence mammary tumour phenotype. Here, we use a CRE-activated HER2 orthologue to specifically target HER2/ERBB2 oncogenic activity to basal or luminal ER− mammary epithelial cells and perform a detailed analysis of the tumours that develop. We find that, in contrast to our previous studies, basal epithelial cells are less sensitive to transformation by the activated NeuKI allele, with mammary epithelial tumour formation largely confined to luminal ER− cells. Histologically, most tumours that developed were classified as either adenocarcinomas of no special type or as metaplastic adenosquamous tumours. The former were typically characterized by amplification of the NeuNT/Erbb2 locus; in contrast, tumours displaying squamous metaplasia were enriched in animals that had been through at least one pregnancy and typically had lower levels of NeuNT/Erbb2 locus amplification but had activated canonical WNT signalling. Squamous changes in these tumours were associated with activation of the epidermal differentiation cluster. Thus, in this model of HER2 breast cancer, cell-of-origin, reproductive history, NeuNT/Erbb2 locus amplification and the activation of specific branches of the WNT signalling pathway all interact to drive inter-tumour heterogeneity., Summary: Using a mouse model of breast cancer, the authors show mammary epithelial cell-type sensitivity to transformation by HER2, as well as a change in tumour phenotype associated with reproductive history and driven by WNT signalling.

Understanding the mechanisms underlying tumour heterogeneity is key to development of treatments that can target specific tumour subtypes. We have previously targeted CRE recombinase-dependent conditional deletion of the tumour suppressor genes Brca1, Brca2, p53 and/or Pten to basal or luminal ER− cells of the mouse mammary epithelium. We demonstrated that both the cell-of-origin and the tumour-initiating genetic lesions co-operate to influence mammary tumour phenotype. Here, we use a CRE-activated HER2 orthologue to specifically target HER2/ERBB2 oncogenic activity to basal or luminal ER− mammary epithelial cells and carry out a detailed analysis of the tumours which develop. We find that in contrast to our previous studies, basal epithelial cells are refractory to transformation by the activated NeuKI allele, with mammary epithelial tumour formation largely confined to luminal ER− cells. Histologically, the majority of tumours that developed were classified as either adenocarcinomas of no special type or metaplastic adenosquamous tumours. Remarkably, the former were more strongly associated with virgin animals and were typically characterised by amplification of the NeuNT/ErbB2 locus and activation of non-canonical WNT signalling. In contrast, tumours characterised by squamous metaplasia were associated with animals that had been through at least one pregnancy and typically had lower levels of NeuNT/ErbB2 locus amplification but had activated canonical WNT signalling. Squamous changes in these tumours were associated with activation of the Epidermal Differentiation Cluster. Thus, in this model of HER2 breast cancer, cell-of-origin, reproductive history, NeuNT/ErbB2 locus amplification, and the activation of specific branches of the WNT signalling pathway all interact to drive inter-tumour heterogeneity.

Purpose: PTEN-null tumors become dependent on the PI3Kβ isoform and can be targeted by molecules such as the selective PI3Kβ inhibitor AZD8186. However, beyond the modulation of the canonical PI3K pathway, the consequences of inhibiting PI3Kβ are poorly defined. Experimental Design: To determine the broader impact of AZD8186 in PTEN-null tumors, we performed a genome-wide RNA-seq analysis of PTEN-null triple-negative breast tumor xenografts treated with AZD8186. Mechanistic consequences of AZD8186 treatment were examined across a number of PTEN-null cell lines and tumor models. Results: AZD8186 treatment resulted in modification of transcript and protein biomarkers associated with cell metabolism. We observed downregulation of cholesterol biosynthesis genes and upregulation of markers associated with metabolic stress. Downregulation of cholesterol biosynthesis proteins, such as HMGCS1, occurred in PTEN-null cell lines and tumor xenografts sensitive to AZD8186. Therapeutic inhibition of PI3Kβ also upregulated PDHK4 and increased PDH phosphorylation, indicative of reduced carbon flux into the TCA cycle. Consistent with this, metabolomic analysis revealed a number of changes in key carbon pathways, nucleotide, and amino acid biosynthesis. Conclusions: This study identifies novel mechanistic biomarkers of PI3Kβ inhibition in PTEN-null tumors supporting the concept that targeting PI3Kβ may exploit a metabolic dependency that contributes to therapeutic benefit in inducing cell stress. Considering these additional pathways will guide biomarker and combination strategies for this class of agents. Clin Cancer Res; 23(24); 7584–95. ©2017 AACR.

Tumor mutational burden (TMB) is emerging as a promising predictive biomarker to select cancer patients who will benefit from immunotherapy. However, accurate TMB estimation from sequencing data remains hampered by the ability to correctly distinguish somatic variants from both germline variants and spurious base changes introduced by sample preparation and sequencing processes. The latter is particularly prevalent in poor quality FFPE material where degradation and chemical modifications of the DNA manifest as variant artefacts leading to over-estimation of TMB. One method of removing germline variants and artefacts is to compare tumor variant calls to those obtained from matched normal tissue, however inclusion of a matched normal adds significant logistical and financial challenges to clinical trial assays. Therefore, there remains a critical need to establish alternative approaches enabling accurate TMB estimation from FFPE tumor-only samples. To address these challenges, we have developed a tumor-only somatic variant calling pipeline for accurate determination of TMB from FFPE material profiled by whole exome sequencing (WES). The process is underpinned by two key features (a) enhanced germline variant filtering that uses a systematic rule-based strategy, (b) FFPE artefact removal using an ensemble machine learning model incorporating multiple variant call QC parameters. The germline filtering approach was developed on a matched fresh frozen (FF)/FFPE/blood dataset of 23 cancer patients. The FFPE artefact filtering model was built on a dataset of 59 FF and FFPE matched pairs, where FFPE artefacts were defined as variants present in FFPE but absent in the FF sample. Benchmarked against internal and TCGA datasets of somatic variant calls from >30 FF samples with a matched normal, our germline and FFPE filtering steps decreased false discovery rates by over 20%. This translated into improved TMB scoring performance with tumor-only pipeline estimates from three SeraCare reference standards closely matching expected TMB scores. Critically, we establish the clinical relevance of our approach on a published cohort of 98 patients treated with an immune checkpoint inhibitor. Here we observed increased association between patient response and TMB estimated from our pipeline (p=0.042) compared to a standard pipeline with no enhanced germline or FFPE artefact filters (p=0.56). Overall, our combined germline and FFPE artefact filtering pipeline results in significantly improved somatic variant detection and TMB scoring performance in FFPE samples where a matched normal is unavailable. It is applicable across multiple disease types and adaptable to other platforms. Therefore, the pipeline has utility in clinical laboratories where accurate somatic variant detection is crucial to successful deployment of predictive biomarkers such as TMB. Citation Format: Hui Sun Leong, Emma O'Connor, Nuala McCabe, Sinead Donegan, Steven M. Walker, Jayna Mistry, Karen Keating, Denis P. Harkin, Richard D. Kennedy, Laura A. Knight, James R. Bradford. A machine learning approach to detect FFPE artefacts leads to accurate estimation of tumor mutational burden from sequencing data without a matched normal [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 152.

Cryptococcus neoformans is an opportunistic fungal infection that causes cryptococcal meningitis in immunocompromised individuals. Macrophages play a critical role in determining the outcome of infection, and can either phagocytose and kill the cryptococcal cells, or disseminate infection. While it is known that macrophages impact the progression of cryptococcal disease, it is not known how the macrophage intracellular niche contributes to complex infection outcomes. Clinical and experimental studies have identified potential genetic differences, in both host and pathogen, but statistical robustness has been difficult to achieve due the large variability in the outcome of infections. Therefore, in attempt to quantitatively explain this variability, we have used a zebrafish model of cryptococcal infection where we can directly relate the initial level of fungal infection with final infection outcome. We find that at low levels of initial fungal burden the outcome of infection is stochastic, while over high ranges of initial infection the outcome is linearly related to the initial burden, but with a further stochastic component that contributes to increased variability. Using these experimental data and data from clinical trials we have generated a computational simulation of in vivo cryptococcal infections which allows us to consider different infection variables and how they alter the progression of cryptococcosis. By combining such computational simulations with our experimental models we demonstrate that the macrophage intracellular niche determines the unknown stochastic component, independent of fungal burden. Using this knowledge, we can better identify the molecular, population genetic, and clinical parameters associated with the outcome of cryptococcosis.

// James R. Bradford 1 , Mark Wappett 2 , Garry Beran 2 , Armelle Logie 2 , Oona Delpuech 2 , Henry Brown 2 , Joanna Boros 2 , Nicola J. Camp 3 , Robert McEwen 2 , Anne Marie Mazzola 4 , Celina D’Cruz 4 , Simon T. Barry 2 1 Department of Oncology and Metabolism, University of Sheffield, Sheffield, South Yorkshire, UK 2 Oncology iMED, AstraZeneca Pharmaceuticals, Alderley Park, Cheshire, UK 3 Department of Internal Medicine and Huntsman Cancer Institute, University of Utah, Salt Lake City, Utah, USA 4 Oncology iMED, AstraZeneca Pharmaceuticals, Gatehouse Park, Massachusetts, USA Correspondence to: James R. Bradford, e-mail: J.R.Bradford@sheffield.ac.uk Keywords: patient-derived xenograft, RNA-Seq, tumor stroma, biomarker discovery, pre-clinical research Received: November 09, 2015 Accepted: February 18, 2016 Published: March 09, 2016 ABSTRACT The tumor microenvironment is emerging as a key regulator of cancer growth and progression, however the exact mechanisms of interaction with the tumor are poorly understood. Whilst the majority of genomic profiling efforts thus far have focused on the tumor, here we investigate RNA-Seq as a hypothesis-free tool to generate independent tumor and stromal biomarkers, and explore tumor-stroma interactions by exploiting the human-murine compartment specificity of patient-derived xenografts (PDX). Across a pan-cancer cohort of 79 PDX models, we determine that mouse stroma can be separated into distinct clusters, each corresponding to a specific stromal cell type. This implies heterogeneous recruitment of mouse stroma to the xenograft independent of tumor type. We then generate cross-species expression networks to recapitulate a known association between tumor epithelial cells and fibroblast activation, and propose a potentially novel relationship between two hypoxia-associated genes, human MIF and mouse Ddx6 . Assessment of disease subtype also reveals MMP12 as a putative stromal marker of triple-negative breast cancer. Finally, we establish that our ability to dissect recruited stroma from trans-differentiated tumor cells is crucial to identifying stem-like poor-prognosis signatures in the tumor compartment. In conclusion, RNA-Seq is a powerful, cost-effective solution to global analysis of human tumor and mouse stroma simultaneously, providing new insights into mouse stromal heterogeneity and compartment-specific disease markers that are otherwise overlooked by alternative technologies. The study represents the first comprehensive analysis of its kind across multiple PDX models, and supports adoption of the approach in pre-clinical drug efficacy studies, and compartment-specific biomarker discovery.

BACKGROUND A substantial number of melanoma patients develop local or metastatic recurrence, and early detection of these is vital to maximise benefit from new therapies such as inhibitors of BRAF and MEK, or immune checkpoints. This study explored the use of novel DNA copy-number profiles in circulating cell-free DNA (cfDNA) as a potential biomarker of active disease and survival. PATIENTS AND METHODS Melanoma patients were recruited from oncology and dermatology clinics in Sheffield, UK, and cfDNA was isolated from stored blood plasma. Using low-coverage whole-genome sequencing, we created copy-number profiles from cfDNA from 83 melanoma patients, 44 of whom had active disease. We used scoring algorithms to summarize copy-number aberrations and investigated their utility in multivariable logistic and Cox regression analyses. RESULTS The copy-number aberration score (CNAS) was a good discriminator of active disease (odds ratio, 3.1; 95% CI, 1.5–6.2; P = 0.002), and CNAS above or below the 75th percentile remained a significant discriminator in multivariable analysis for active disease (P = 0.019, with area under ROC curve of 0.90). Additionally, mortality was higher in those with CNASs above the 75th percentile than in those with lower scores (HR, 3.4; 95% CI, 1.5–7.9; P = 0.005), adjusting for stage of disease, disease status (active or resected), BRAF status, and cfDNA concentration. CONCLUSIONS This study demonstrates the potential of a de novo approach utilizing copy-number profiling of cfDNA as a biomarker of active disease and survival in melanoma. Longitudinal analysis of copy-number profiles as an early marker of relapsed disease is warranted.

Background The incidence of human papilloma virus positive (HPV+) oropharyngeal squamous cell carcinoma (OPSCC) has increased rapidly in recent decades. These tumours have a favourable outcome compared to HPV‐negative (HPV−) OPSCC. However, HPV+ tumours are more likely to metastasise to distant sites, suggesting a difference in how these tumour subtypes interact with the metastatic niche. Extracellular vesicles (EVs) have emerged as important players in cell‐to‐cell communication and are a potential source of biomarkers for cancer diagnosis. This study aims to characterise the microRNA cargo of small EVs released by HPV+ and HPV− OPSCC cell lines. Methods Extracellular vesicles produced by HPV+ (SCC2 and SCC90) and HPV− (SCC72 an SCC89) OPSCC cells were characterised by tunable resistive pulse sensing (TRPS) and western blotting. RNA was extracted from EVs and analysed by small RNA sequencing. A bioinformatics approach was used to identify EV miRNA signatures associated with HPV status. Results HPV− OPSCC cells produced more EVs than HPV+ OPSCC cells. EVs were positive for the common EV markers CD63, CD9 and TSG101. Unbiased hierarchical clustering analysis of EV miRNA cargo revealed that samples clustered based on HPV status. 14 miRNA were enriched in HPV+ cell‐derived EVs, whereas 19 miRNA were enriched in EVs derived from HPV− cell lines. Conclusions Here, we identify EV miRNA signatures indicative of the HPV status of the parent cell. This may provide a platform from which to validate salivary or blood‐based biomarkers with utility for early detection and stratifying risk in OPSCC patients.

Long non-coding RNAs (lncRNAs) are emerging as crucial regulators of cellular processes and diseases such as cancer; however, their functions remain poorly characterised. Several studies have demonstrated that lncRNAs are typically disease and tumour subtype specific, particularly in breast cancer where lncRNA expression alone is sufficient to discriminate samples based on hormone status and molecular intrinsic subtype. However, little attempt has been made to assess the reproducibility of lncRNA signatures across more than one dataset. In this work, we derive consensus lncRNA signatures indicative of breast cancer subtype based on two clinical RNA-Seq datasets: the Utah Breast Cancer Study and The Cancer Genome Atlas, through integration of differential expression and hypothesis-free clustering analyses. The most consistent signature is associated with breast cancers of the basal-like subtype, leading us to generate a putative set of six lncRNA basal-like breast cancer markers, at least two of which may have a role in cis-regulation of known poor prognosis markers. Through in silico functional characterization of individual signatures and integration of expression data from pre-clinical cancer models, we discover that discordance between signatures derived from different clinical cohorts can arise from the strong influence of non-cancerous cells in tumour samples. As a consequence, we identify nine lncRNAs putatively associated with breast cancer associated fibroblasts, or the immune response. Overall, our study establishes the confounding effects of tumour purity on lncRNA signature derivation, and generates several novel hypotheses on the role of lncRNAs in basal-like breast cancers and the tumour microenvironment.

Long non-coding RNAs (lncRNAs) are emerging as crucial regulators of cellular processes in diseases such as cancer, although the functions of most remain poorly understood. To address this, here we apply a novel strategy to integrate gene expression profiles across 32 cancer types, and cluster human lncRNAs based on their pan-cancer protein-coding gene associations. By doing so, we derive 16 lncRNA modules whose unique properties allow simultaneous inference of function, disease specificity and regulation for over 800 lncRNAs. Remarkably, modules could be grouped into just four functional themes: transcription regulation, immunological, extracellular, and neurological, with module generation frequently driven by lncRNA tissue specificity. Notably, three modules associated with the extracellular matrix represented potential networks of lncRNAs regulating key events in tumour progression. These included a tumour-specific signature of 33 lncRNAs that may play a role in inducing epithelialmesenchymal transition through modulation of TGFβ signalling, and two stromal-specific modules comprising 26 lncRNAs linked to a tumour suppressive microenvironment, and 12 lncRNAs related to cancer-associated fibroblasts. At least one member of the 12-lncRNA signature was experimentally supported by siRNA knockdown, which resulted in attenuated differentiation of quiescent fibroblasts to a cancer-associated phenotype. Overall, the study provides a unique pan-cancer perspective on the lncRNA functional landscape, acting as a global source of novel hypotheses on lncRNA contribution to tumour progression.Author SummaryThe established view of protein production is that genomic DNA is transcribed into RNA, which is then translated into protein. Proteins play a critical role in shaping the function of each individual cell in the human body yet they represent less than 2% of human genomic sequence whilst up to 90% of the genome is transcribed. To explain this disparity, the existence of thousands of long non-coding RNAs (lncRNAs) has emerged that do not encode proteins but perform function as an RNA molecule. Most lncRNAs have yet to be assigned a specific biological role, so to address this we apply a novel computational approach to characterise the function of >800 lncRNAs through consistent association with protein coding genes across multiple cancer types. By doing so, we discover 16 “modules” of closely related lncRNAs that share broad functional themes, the most compelling of which consists of 12 lncRNAs that could regulate activation of specific cells neighbouring the tumour, leading to accelerated tumour progression and invasion. Overall, the study provides the most robust view of the lncRNA-protein coding gene landscape to date, adding to growing evidence that lncRNAs are key regulators of cancer, and have therapeutic potential comparable to proteins.

Prostate cancer research is hampered by the lack of in vivo preclinical models that accurately reflect patient tumour biology and the clinical heterogeneity of human prostate cancer. To overcome these limitations we propagated and characterised a new collection of patient-derived prostate cancer xenografts. Tumour fragments from 147 unsupervised, surgical prostate samples were implanted subcutaneously into immunodeficient Rag2(-/-)gamma C-/- mice within 24 hours of surgery. Histologic and molecular characterisation of xenografts was compared with patient characteristics, including androgen-deprivation therapy, and exome sequencing. Xenografts were established from 47 of 147 (32%) implanted primary prostate cancers. Only 14% passaged successfully resulting in 20 stable lines; derived from 20 independent patient samples. Surprisingly, only three of the 20 lines (15%) were confirmed as prostate cancer; one line comprised of mouse stroma, and 16 were verified as human donor-derived lymphoid neoplasms. PCR for Epstein-Barr Virus (EBV) nuclear antigen, together with exome sequencing revealed that the lymphomas were exclusively EBV-associated. Genomic analysis determined that 14 of the 16 EBV+ lines had unique monoclonal or oligoclonal immunoglobulin heavy chain gene rearrangements, confirming their B-cell origin. We conclude that the generation of xenografts from tumour fragments can commonly result in B-cell lymphoma from patients carrying latent EBV. We recommend routine screening, of primary outgrowths, for latent EBV to avoid this phenomenon.

e21045 Background: Traditionally gene expression signatures (GES) are used individually to classify patients into subgroups. Signatures targeting the same biology are often developed independently and may not classify identically. We developed the claraT software tool that uses consensus between multiple published GES categorised by the Hallmarks of Cancer (Hanahan & Weinberg, 2011) to classify cancers. As metastatic melanoma represents poor prognostic disease (5-yr survival 15-20%), we applied claraT to the TCGA melanoma dataset to identify targetable biologies, validated in a cohort of melanoma patients treated with Ipilimumab. Methods: TCGA RNA-seq data ( n= 472) was analysed using the claraT platform including GES for immune ( n= 14), angiogenesis ( n= 9) and epithelial-mesenchymal transition (EMT) ( n= 12) Hallmarks. Samples were clustered for the combined and individual Hallmarks. Median progression-free (PFS) and overall-survival (OS) differences were analysed across identified subgroups. Analysis was validated in an Ipilimumab treated melanoma dataset ( n= 42) (Van Allen, 2015). Results: Clustering the combined Hallmarks identified 4 subgroups in the TCGA cohort: 1) Immune active, 2) Immune-EMT active, 3) EMT-Angiogenesis active, 4) All inactive. Groups 1&2 had significantly improved OS compared to Groups 3&4 (HR = 0.50, p< 0.0001). Clustering using single Hallmarks revealed that immune-positive tumours had significantly improved OS (HR = 0.53, p< 0.0001) compared to immune-negative tumours. Angiogenesis-negative tumours displayed improved PFS (HR = 0.73, p= 0.03) and OS (HR = 0.53, p

Molecular profiling using low coverage whole genome sequencing of cell free DNA (cfDNA) represents a non-targeted approach to identify multiple somatic copy number alterations (SCNA) across different lung cancer subtypes. We aim to establish that SCNA can be detected in cfDNA of lung cancer cases.

Background Identification of synthetic lethal interactions in cancer cells could offer promising new therapeutic targets. Large-scale functional genomic screening presents an opportunity to test large numbers of cancer synthetic lethal hypotheses. Methods enriching for candidate synthetic lethal targets in molecularly defined cancer cell lines can steer effective design of screening efforts. Loss of one partner of a synthetic lethal gene pair creates a dependency on the other, thus synthetic lethal gene pairs should never show simultaneous loss-of-function. We have developed a computational approach to mine large multi-omic cancer data sets and identify gene pairs with mutually exclusive loss-of-function. Since loss-of-function may not always be genetic, we look for deleterious mutations, gene deletion and/or loss of mRNA expression by bimodality defined with a novel algorithm BiSEp. Results Applying this toolkit to both tumour cell line and patient data, we achieve statistically significant enrichment for experimentally validated tumour suppressor genes and synthetic lethal gene pairings. Notably non-reliance on genetic loss reveals a number of known synthetic lethal relationships otherwise missed, resulting in marked improvement over genetic-only predictions. We go on to establish biological rationale surrounding a number of novel candidate synthetic lethal gene pairs with demonstrated dependencies in published cancer cell line shRNA screens. Conclusions This work introduces a multi-omic approach to define gene loss-of-function, and enrich for candidate synthetic lethal gene pairs in cell lines testable through functional screens. In doing so, we offer an additional resource to generate new cancer drug target and combination hypotheses. Algorithms discussed are freely available in the BiSEp CRAN package at http://cran.r-project.org/web/packages/BiSEp/index.html. Electronic supplementary material The online version of this article (doi:10.1186/s12864-016-2375-1) contains supplementary material, which is available to authorized users.

Motivation: Functional genomics data provides a rich source of information that can be used in the annotation of the thousands of genes of unknown function found in most sequenced genomes. However, previous gene function prediction programs are mostly produced for relatively well-annotated organisms that often have a large amount of functional genomics data. Here, we present a novel method for predicting gene function that uses clustering of genes by semantic similarity, a naïve Bayes classifier and ‘enrichment analysis’ to predict gene function for a genome that is less well annotated but does has a severe effect on human health, that of the malaria parasite Plasmodium falciparum. Results: Predictions for the molecular function, biological process and cellular component of P.falciparum genes were created from eight different datasets with a combined prediction also being produced. The high-confidence predictions produced by the combined prediction were compared to those produced by a simple K-nearest neighbour classifier approach and were shown to improve accuracy and coverage. Finally, two case studies are described, which investigate two biological processes in more detail, that of translation initiation and invasion of the host cell. Availability: Predictions produced are available at http://www.bioinformatics.leeds.ac.uk/∼bio5pmrt/PAGODA Contact: D.R.Westhead@leeds.ac.uk Supplementary information: Supplementary data are available at Bioinformatics online.

*† ‡ SUMMARY Despite recent advances, accurate gene function prediction remains an elusive goal, with very few methods directly applicable to the plant Arabidopsis thaliana. In this study, we present GO-At (gene ontology prediction in A. thaliana), a method that combines five data types (co-expression, sequence, phylogenetic profile, interaction and gene neighbourhood) to predict gene function in Arabidopsis. Using a simple, yet powerful two-step approach, GO-At first generates a list of genes ranked in descending order of probability of functional association with the query gene. Next, a prediction score is automatically assigned to each function in this list based on the assumption that functions appearing most frequently at the top of the list are most likely to represent the function of the query gene. In this way, the second step provides an effective alternative to simply taking the ‘best hit’ from the first list, and achieves success rates of up to 79%. GO-At is applicable across all three GO categories: molecular function, biological process and cellular component, and can assign functions at multiple levels of annotation detail. Furthermore, we demonstrate GO-At’s ability to predict functions of uncharacterized genes by identifying ten putative golgins/Golgi-associated proteins amongst 8219 genes of previously unknown cellular component and present independent evidence to support our predictions. A web-based implementation of GO-At (http://www.bioinformatics.leeds.ac.uk/goat) is available, providing a unique resource for plant researchers to make predictions for uncharacterized genes and predict novel functions in Arabidopsis.

A method has been developed for the prediction of proteins involved in genetic disorders. This involved combining deleterious SNP prediction with a system based on protein interactions and phenotype distances; this is the first time that deleterious SNP prediction has been used to make predictions across linkage-intervals. At each step we tested and selected the best procedure, revealing that the computationally expensive method of assigning medical meta-terms to create a phenotype distance matrix was outperformed by a simple word counting technique. We carried out in-depth benchmarking with increasingly stringent data sets, reaching precision values of up to 75% (19% recall) for 10-Mb linkage-intervals (averaging 100 genes). For the most stringent (worst-case) data we attained an overall recall of 6%, yet still achieved precision values of up to 90% (4% recall). At all levels of stringency and precision the addition of predicted deleterious SNPs was shown to increase recall. Hum Mutat 0, 1–9, 2009. © 2009 Wiley-Liss, Inc.

Although endocrine therapy is successfully used to treat patients with estrogen receptor (ER) positive breast cancer, a substantial proportion of this population will relapse. Several mechanisms of acquired resistance have been described including activation of the mTOR pathway, increased activity of CDK4 and activating mutations in ER. Using a patient derived xenograft model harboring a common activating ER ligand binding domain mutation (D538G), we evaluated several combinatorial strategies using the selective estrogen receptor degrader (SERD) fulvestrant in combination with chromatin modifying agents, and CDK4/6 and mTOR inhibitors. In this model, fulvestrant binds WT and MT ER, reduces ER protein levels, and downregulated ER target gene expression. Addition of JQ1 or vorinostat to fulvestrant resulted in tumor regression (41% and 22% regression, respectively) though no efficacy was seen when either agent was given alone. Interestingly, although the CDK4/6 inhibitor palbociclib and mTOR inhibitor everolimus were efficacious as monotherapies, long-term delayed tumor growth was only observed when co-administered with fulvestrant. This observation was consistent with a greater inhibition of compensatory signaling when palbociclib and everolimus were co-dosed with fulvestrant. The addition of fulvestrant to JQ1, vorinostat, everolimus and palbociclib also significantly reduced lung metastatic burden as compared to monotherapy. The combination potential of fulvestrant with palbociclib or everolimus were confirmed in an MCF7 CRISPR model harboring the Y537S ER activating mutation. Taken together, these data suggest that fulvestrant may have an important role in the treatment of ER positive breast cancer with acquired ER mutations.

Motivation: Structural genomics projects are beginning to produce protein structures with unknown function, therefore, accurate, automated predictors of protein function are required if all these structures are to be properly annotated in reasonable time. Identifying the interface between two interacting proteins provides important clues to the function of a protein and can reduce the search space required by docking algorithms to predict the structures of complexes. Results: We have combined a support vector machine (SVM) approach with surface patch analysis to predict protein--protein binding sites. Using a leave-one-out cross-validation procedure, we were able to successfully predict the location of the binding site on 76% of our dataset made up of proteins with both transient and obligate interfaces. With heterogeneous cross-validation, where we trained the SVM on transient complexes to predict on obligate complexes (and vice versa), we still achieved comparable success rates to the leave-one-out cross-validation suggesting that sufficient properties are shared between transient and obligate interfaces. Availability: A web application based on the method can be found at http://www.bioinformatics.leeds.ac.uk/ppi_pred. The dataset of 180 proteins used in this study is also available via the same web site. Contact: westhead@bmb.leeds.ac.uk Supplementary information: http://www.bioinformatics.leeds.ac.uk/ppi-pred/supp-material

We have carried out a thorough and systematic sequence-structure study on how the pattern of conservation at the interface differs from the noninteracting surface in seven proteases and their inhibitors. As expected, the interface of a protease could be easily distinguished from the noninteracting surface by a concentrated area of conservation. In contrast, there was less distinction to be made between the interface and the noninteracting surface of inhibitors, and in five of the seven cases, a higher proportion of the interface area was variable compared to the rest of the surface. This is likely to cause a problem for binding-site prediction methods that assume the largest cluster of highly conserved residues on the surface of a protein corresponds to the interface. We conclude that such methods would succeed when applied to our protease test cases, but complications could arise with the inhibitors. These results also impact on methods to solve the protein-protein docking problem that use conservation at the interface to provide the location of the two protein binding sites prior to application of the docking algorithm.

Bayesian networks are increasingly important for integrating biological data and for inferring cellular networks and pathways. What are Bayesian networks and how are they used for inference?

Pre-clinical models of tumour biology often rely on propagating human tumour cells in a mouse. In order to gain insight into the alignment of these models to human disease segments or investigate the effects of different therapeutics, approaches such as PCR or array based expression profiling are often employed despite suffering from biased transcript coverage, and a requirement for specialist experimental protocols to separate tumour and host signals. Here, we describe a computational strategy to profile transcript expression in both the tumour and host compartments of pre-clinical xenograft models from the same RNA sample using RNA-Seq. Key to this strategy is a species-specific mapping approach that removes the need for manipulation of the RNA population, customised sequencing protocols, or prior knowledge of the species component ratio. The method demonstrates comparable performance to species-specific RT-qPCR and a standard microarray platform, and allowed us to quantify gene expression changes in both the tumour and host tissue following treatment with cediranib, a potent vascular endothelial growth factor receptor tyrosine kinase inhibitor, including the reduction of multiple murine transcripts associated with endothelium or vessels, and an increase in genes associated with the inflammatory response in response to cediranib. In the human compartment, we observed a robust induction of hypoxia genes and a reduction in cell cycle associated transcripts. In conclusion, the study establishes that RNA-Seq can be applied to pre-clinical models to gain deeper understanding of model characteristics and compound mechanism of action, and to identify both tumour and host biomarkers.

Strand-specific RNA sequencing of S. pombe reveals a highly structured programme of ncRNA expression at over 600 loci. Functional investigations show that this extensive ncRNA landscape controls the complex programme of sexual differentiation in S. pombe., The model eukaryote S. pombe features substantial numbers of ncRNAs many of which are antisense regulatory transcripts (ARTs), ncRNAs expressed on the opposing strand to coding sequences. Individual ARTs are generated during the mitotic cycle, or at discrete stages of sexual differentiation to downregulate the levels of proteins that drive and coordinate sexual differentiation. Antisense transcription occurring from events such as bidirectional transcription is not simply artefactual ‘chatter', it performs a critical role in regulating gene expression., Regulation of the RNA profile is a principal control driving sexual differentiation in the fission yeast Schizosaccharomyces pombe. Before transcription, RNAi-mediated formation of heterochromatin is used to suppress expression, while post-transcription, regulation is achieved via the active stabilisation or destruction of transcripts, and through at least two distinct types of splicing control (Mata et al, 2002; Shimoseki and Shimoda, 2001; Averbeck et al, 2005; Mata and Bähler, 2006; Xue-Franzen et al, 2006; Moldon et al, 2008; Djupedal et al, 2009; Amorim et al, 2010; Grewal, 2010; Cremona et al, 2011). Around 94% of the S. pombe genome is transcribed (Wilhelm et al, 2008). While many of these transcripts encode proteins (Wood et al, 2002; Bitton et al, 2011), the majority have no known function. We used a strand-specific protocol to sequence total RNA extracts taken from vegetatively growing cells, and at different points during a time course of sexual differentiation. The resulting data redefined existing gene coordinates and identified additional transcribed loci. The frequency of reads at each of these was used to monitor transcript abundance. Transcript levels at 6599 loci changed in at least one sample (G-statistic; False Discovery Rate, Strand-specific RNA sequencing of S. pombe revealed a highly structured programme of ncRNA expression at over 600 loci. Waves of antisense transcription accompanied sexual differentiation. A substantial proportion of ncRNA arose from mechanisms previously considered to be largely artefactual, including improper 3′ termination and bidirectional transcription. Constitutive induction of the entire spk1+, spo4+, dis1+ and spo6+ antisense transcripts from an integrated, ectopic, locus disrupted their respective meiotic functions. This ability of antisense transcripts to disrupt gene function when expressed in trans suggests that cis production at native loci during sexual differentiation may also control gene function. Consistently, insertion of a marker gene adjacent to the dis1+ antisense start site mimicked ectopic antisense expression in reducing the levels of this microtubule regulator and abolishing the microtubule-dependent ‘horsetail' stage of meiosis. Antisense production had no impact at any of these loci when the RNA interference (RNAi) machinery was removed. Thus, far from being simply ‘genome chatter', this extensive ncRNA landscape constitutes a fundamental component in the controls that drive the complex programme of sexual differentiation in S. pombe.

Background RNA-Seq exploits the rapid generation of gigabases of sequence data by Massively Parallel Nucleotide Sequencing, allowing for the mapping and digital quantification of whole transcriptomes. Whilst previous comparisons between RNA-Seq and microarrays have been performed at the level of gene expression, in this study we adopt a more fine-grained approach. Using RNA samples from a normal human breast epithelial cell line (MCF-10a) and a breast cancer cell line (MCF-7), we present a comprehensive comparison between RNA-Seq data generated on the Applied Biosystems SOLiD platform and data from Affymetrix Exon 1.0ST arrays. The use of Exon arrays makes it possible to assess the performance of RNA-Seq in two key areas: detection of expression at the granularity of individual exons, and discovery of transcription outside annotated loci. Results We found a high degree of correspondence between the two platforms in terms of exon-level fold changes and detection. For example, over 80% of exons detected as expressed in RNA-Seq were also detected on the Exon array, and 91% of exons flagged as changing from Absent to Present on at least one platform had fold-changes in the same direction. The greatest detection correspondence was seen when the read count threshold at which to flag exons Absent in the SOLiD data was set to t Conclusions By focusing on exon-level expression, we present the most fine-grained comparison between RNA-Seq and microarrays to date. Overall, our study demonstrates that data from a SOLiD RNA-Seq experiment are sufficient to generate results comparable to those produced from Affymetrix Exon arrays, even using only a single replicate from each platform, and when presented with a large genome.

The transcriptional repressor B lymphocyte-induced maturation protein-1 (BLIMP1) regulates gene expression and cell fate. The DNA motif bound by BLIMP1 in vitro overlaps with that of interferon regulatory factors (IRFs), which respond to inflammatory/immune signals. At such sites, BLIMP1 and IRFs can antagonistically regulate promoter activity. In vitro motif selection predicts that only a subset of BLIMP1 or IRF sites is subject to antagonistic regulation, but the extent to which antagonism occurs is unknown, since an unbiased assessment of BLIMP1 occupancy in vivo is lacking. To address this, we identified an extended set of promoters occupied by BLIMP1. Motif discovery and enrichment analysis demonstrate that multiple motif variants are required to capture BLIMP1 binding specificity. These are differentially associated with CpG content, leading to the observation that BLIMP1 DNA-binding is methylation sensitive. In occupied promoters, only a subset of BLIMP1 motifs overlap with IRF motifs. Conversely, a distinct subset of IRF motifs is not enriched amongst occupied promoters. Genes linked to occupied promoters containing overlapping BLIMP1/IRF motifs (e.g. AIM2, SP110, BTN3A3) are shown to constitute a dynamic target set which is preferentially activated by BLIMP1 knock-down. These data confirm and extend the competitive model of BLIMP1 and IRF interaction.

The genome sequence of the malaria parasite Plasmodium falciparum was published in 2002 and revealed that approximately 60% of its genes could not be assigned a function. Eight years later the majority of P. falciparum proteins are still of unknown function. We therefore present PlasmoPredict, an easy-to-use online gene function prediction tool that integrates a wide range of functional genomics data for P. falciparum to aid in the annotation of these genes.

Malaria is one of the world's most deadly diseases and is caused by the parasite Plasmodium falciparum . Sixty percent of P. falciparum genes have no known function and therefore new methods of gene function prediction are needed. To address this problem, we train a naive Bayes classifier on multiple sources of data and subsequently apply a modified version of the Gene Set Enrichment Analysis Algorithm to predict gene function in P. falciparum . To define gene function, we exploit the hierarchical structure of the Gene Ontology, specifically using the Biological Process category. We demonstrate the value of integrating multiple data sources by achieving accurate predictions on genes that cannot be annotated using simple sequence similarity based methods.

Bayesian networks (BNs) provide a neat and compact representation for expressing joint probability distributions (JPDs) and for inference. They are becoming increasingly important in the biological sciences for the tasks of inferring cellular networks [1], modelling protein signalling pathways [2], systems biology, data integration [3], classification [4], and genetic data analysis [5]. The representation and use of probability theory makes BNs suitable for combining domain knowledge and data, expressing causal relationships, avoiding overfitting a model to training data, and learning from incomplete datasets. The probabilistic formalism provides a natural treatment for the stochastic nature of biological systems and measurements. This primer aims to introduce BNs to the computational biologist, focusing on the concepts behind methods for learning the parameters and structure of models, at a time when they are becoming the machine learning method of choice. There are many applications in biology where we wish to classify data; for example, gene function prediction. To solve such problems, a set of rules are required that can be used for prediction, but often such knowledge is unavailable, or in practice there turn out to be many exceptions to the rules or so many rules that this approach produces poor results. Machine learning approaches often produce better results, where a large number of examples (the training set) is used to adapt the parameters of a model that can then be used for performing predictions or classifications on data. There are many different types of models that may be required and many different approaches to training the models, each with its pros and cons. An excellent overview of the topic can be found in [6] and [7]. Neural networks, for example, are often able to learn a model from training data, but it is often difficult to extract information about the model, which with other methods can provide valuable insights into the data or problem being solved. A common problem in machine learning is overfitting, where the learned model is too complex and generalises poorly to unseen data. Increasing the size of the training dataset may reduce this; however, this assumes more training data is readily available, which is often not the case. In addition, often it is important to determine the uncertainty in the learned model parameters or even in the choice of model. This primer focuses on the use of BNs, which offer a solution to these issues. The use of Bayesian probability theory provides mechanisms for describing uncertainty and for adapting the number of parameters to the size of the data. Using a graphical representation provides a simple way to visualise the structure of a model. Inspection of models can provide valuable insights into the properties of the data and allow new models to be produced.

The Arabidopsis Co-expression Tool, ACT, ranks the genes across a large microarray dataset according to how closely their expression follows the expression of a query gene. A database stores pre-calculated co-expression results for approximately 21,800 genes based on data from over 300 arrays. These results can be corroborated by calculation of co-expression results for user-defined sub-sets of arrays or experiments from the NASC/GARNet array dataset. Clique Finder (CF) identifies groups of genes which are consistently co-expressed with each other across a user-defined co-expression list. The parameters can be altered easily to adjust cluster size and the output examined for optimal inclusion of genes with known biological roles. Alternatively, a Scatter Plot tool displays the correlation coefficients for all genes against two user-selected queries on a scatter plot which can be useful for visual identification of clusters of genes with similar r-values. User-input groups of genes can be highlighted on the scatter plots. Inclusion of genes with known biology in sets of genes identified using CF and Scatter Plot tools allows inferences to be made about the roles of the other genes in the set and both tools can therefore be used to generate short lists of genes for further characterization. ACT is freely available at www.Arabidopsis.leeds.ac.uk/ACT.

Identifying the interface between two interacting proteins provides important clues to the function of a protein, and is becoming increasing relevant to drug discovery. Here, surface patch analysis was combined with a Bayesian network to predict protein-protein binding sites with a success rate of 82% on a benchmark dataset of 180 proteins, improving by 6% on previous work and well above the 36% that would be achieved by a random method. A comparable success rate was achieved even when evolutionary information was missing, a further improvement on our previous method which was unable to handle incomplete data automatically. In a case study of the Mog1p family, we showed that our Bayesian network method can aid the prediction of previously uncharacterised binding sites and provide important clues to protein function. On Mog1p itself a putative binding site involved in the SLN1-SKN7 signal transduction pathway was detected, as was a Ran binding site, previously characterized solely by conservation studies, even though our automated method operated without using homologous proteins. On the remaining members of the family (two structural genomics targets, and a protein involved in the photosystem II complex in higher plants) we identified novel binding sites with little correspondence to those on Mog1p. These results suggest that members of the Mog1p family bind to different proteins and probably have different functions despite sharing the same overall fold. We also demonstrated the applicability of our method to drug discovery efforts by successfully locating a number of binding sites involved in the protein-protein interaction network of papilloma virus infection. In a separate study, we attempted to distinguish between the two types of binding site, obligate and non-obligate, within our dataset using a second Bayesian network. This proved difficult although some separation was achieved on the basis of patch size, electrostatic potential and conservation. Such was the similarity between the two interacting patch types, we were able to use obligate binding site properties to predict the location of non-obligate binding sites and vice versa.

A single-location, challenge-model study was conducted to evaluate the effectiveness of lincomycin against porcine proliferative enteropathy when administered through the drinking water at 125 and 250 mg/gallon. The primary variables of interest were pig removal rate, diarrhea scores, demeanor scores, and abdominal appearance scores. Ancillary performance variables examined included average daily feed intake, average daily gain, and feed per gain. After a 3-day acclimation period, pigs were challenged on 2 consecutive days with a mucosal homogenate containing a total dose of 1.4 x 10(9) cells of Lawsonia intracellularis. Five days later, when porcine proliferative enteropathy was well established, drinking water medicated with 125 mg (L125) or 250 mg (L250) lincomycin/gallon was provided to two groups of pigs for 10 days. Pigs were observed for 13 days following the treatment period. A third group of pigs served as controls and received unmedicated drinking water throughout the study. The L250 group experienced a significantly lower (P.05) pig removal rate than the control group over the 23-day observation period. Additionally, for every primary variable, the L250 group experienced a significantly decreased (P.01) number of abnormal days compared with the control group. The L125 group showed a significant reduction (P.05) in abnormal demeanor and abnormal abdominal appearance scores compared with controls.

This chapter describes the steps involved in the formation of three-dimensional structure from amino acid sequence. Proteins have adapted their three-dimensional structures to carry out metabolic and regulatory processes such as enzyme catalysis, molecular transport, signaling, the immune response, and gene expression. To understand the relationship between a protein's structure and function fully, knowledge of that protein's evolutionary history and its interactions is essential. Therefore, we describe the mechanisms by which proteins have evolved. We also describe how molecular recognition drives protein interactions and the types of interaction that take place. Keywords: amino acids; protein structure; function; evolution; interactions; molecular recognition

Olaparib is an oral inhibitor of the Poly-(ADP-ribose)-polymerases (PARP-1, -2 and -3), and has demonstrated clinical activity in trials of patients with BRCA-deficient tumors. The BRCA1 and BRCA2 proteins are both important for the repair of DNA double strand breaks (DSBs) by homologous recombination repair (HRR) and sensitivity to olaparib has further been shown preclinically to extend to other DNA DSB repair factor deficiencies, including those in the ataxia telangiectasia mutated (ATM) protein. We have demonstrated that a high proportion of gastric cancer (GC) cell lines (∼50%) are responsive (IC50 Based on these data, a Phase II GC trial has been carried out that enriches for patients with tumors having low or undetectable ATM IHC scores (H-score equivalent less than or equal to 10). Data from this clinical trial confirm activity (overall survival [OS] benefit HR=0.56, 95% CI: 0.35-0.87; P=0.010) in the overall population, which includes ATM tumor positive patients, and importantly, a substantially better OS benefit in the ATM-negative population (HR=0.35, 95% CI: 0.17-0.71; P=0.003). Together, these data provide an important example of a PARP inhibitor demonstrating both preclinical and clinical activity in a well defined but non-BRCA DNA repair deficient background. Citation Format: Darren Hodgson, Helen Mason, Lenka Oplustilova, Chris Harbron, Xiaolu Yin, Seock-Ah Im, Helen Jones, Lai Zhongwu, Brian Dougherty, Matthew McLoughlin, James Bradford, Andrew Dickinson, Anitra Fielding, Jane Robertson, Woo-Ho Kim, Chris Womack, Yi Gu, Yung-Jue Bang, Alan Lau, J. Carl Barrett, Mark J. O'Connor. Activity of the PARP inhibitor olaparib in ATM-deficient gastric cancer: from preclinical models to the clinic. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 2398. doi:10.1158/1538-7445.AM2014-2398

Building upon the exciting precedent exemplified by olaparib in the PARP-BRCA axis, we aim to identify and exploit further synthetic lethal (SL) interactions within and beyond the DNA damage response (DDR) area. To facilitate with this search, we have developed a method implemented as an in silico algorithm that uses gene expression and mutation data to detect putative SL interactions in cell lines and cancer tissue across The Cancer Genome Atlas (TCGA). This approach has detected relatively rare signatures including those described in the literature, for example ATM and ATR. The algorithm has also revealed potentially novel in silico SL gene pairs outside of DDR involving frequently mutated tumor suppressor genes linked to targets that AstraZeneca has previously developed selective compounds against. We are validating an attractive selection of these in vitro using appropriate cell lines, siRNA and small molecule inhibitors. Citation Format: Mark Wappett, Richard Ward, Paul Fisher, James Bradford, Jonathan Dry, Mark O'Connor, Alan Lau, Steve Durant. Exploiting synthetic lethal interactions in the DNA damage response area and beyond in drug discovery. [abstract]. In: Proceedings of the AACR Precision Medicine Series: Synthetic Lethal Approaches to Cancer Vulnerabilities; May 17-20, 2013; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(5 Suppl):Abstract nr A41.

The growth of the majority of prostate tumors is driven by androgens signaling through the androgen receptor. Most patients initially respond to androgen deprivation therapies such as LHRH analogues and anti-androgens like bicalutamide, but after a varying duration of treatment, patients generally develop resistance to these therapies and their tumor progresses to an androgen independent state. Molecular mechanisms that drive androgen independent disease include mutation and amplification of androgen receptor (AR), androgen-independent activation of AR, and AR-independent mechanisms. The aim of this project was to sequence the entire genome of pairs of cell lines, modeling the progression to androgen independent disease, to increase our understanding of these mechanisms and identify new therapeutic targets. Four ‘parental’ prostate cancer cell lines, all of which are sensitive to hormone deprivation, and an androgen independent derivative of each line, derived either by culturing in bicalutamide or growing in the absence of androgens, were subjected to whole genome sequencing. Genomic variants present in the androgen independent but not the parental lines, presumed to be acquired during the progression to androgen independence, were then identified. To find any commonality between the four different pairings, the derived variants were then mapped to functional biological pathways and their likely impact on protein function predicted. Interestingly, derived mutations in the PI3K and Notch signaling pathways were identified in multiple cell lines. In addition to the analysis of derived mutations, baseline characterization of genomic rearrangements and copy number changes in cell lines highlighted variability in the patterns and prevalence of these events as well as uncovering aberrations in genes known to impact prostate cancer development (PTEN, AR, TMPRSS2 rearrangement). This extensive genome sequencing effort has yielded a deeper understanding of potential pathways and molecular events that occur during the acquisition of resistance to anti-androgen therapy, using in vitro models of prostate cancer. These candidates now require validation in tissue samples from prostate cancer patients and functional assays. Citation Format: Gillian L. Dalgliesh, Joanne L. Hartley, Cary H. O'Donnell, James R. Bradford, Iain McKendrick, Barry R. Davies. Whole-genome sequencing of paired androgen-dependent and -independent prostate cancer cell lines: Potential role of novel targets in PI3K and Notch signaling pathways in resistance to hormonal therapy [abstract]. In: Proceedings of the AACR Special Conference on Advances in Prostate Cancer Research; 2012 Feb 6-9; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2012;72(4 Suppl):Abstract nr C38.

Background A number of methods that use both protein structural and evolutionary information are available to predict the functional consequences of missense mutations. However, many of these methods break down if either one of the two types of data are missing. Furthermore, there is a lack of rigorous assessment of how important the different factors are to prediction. Results Here we use Bayesian networks to predict whether or not a missense mutation will affect the function of the protein. Bayesian networks provide a concise representation for inferring models from data, and are known to generalise well to new data. More importantly, they can handle the noisy, incomplete and uncertain nature of biological data. Our Bayesian network achieved comparable performance with previous machine learning methods. The predictive performance of learned model structures was no better than a naïve Bayes classifier. However, analysis of the posterior distribution of model structures allows biologically meaningful interpretation of relationships between the input variables. Conclusion The ability of the Bayesian network to make predictions when only structural or evolutionary data was observed allowed us to conclude that structural information is a significantly better predictor of the functional consequences of a missense mutation than evolutionary information, for the dataset used. Analysis of the posterior distribution of model structures revealed that the top three strongest connections with the class node all involved structural nodes. With this in mind, we derived a simplified Bayesian network that used just these three structural descriptors, with comparable performance to that of an all node network.

Neonatal rats of the Holtzman strain, 6 days of age, were fed a myo-inositol restricted liquid formula by gastric intubation for 10 days, after which they were fed a purified myo-inositol-free diet until they were 72 days old. No differences in weight gain were observed between myo-inositol/100 ml of formula or 150 mg myo-inositol/100 g diet. Most tissues examined from rats fed the myo-inositol deprived formula and diet had lower free myo-inositol levels than the controls with the exception of the liver. Despite reduced free and lipid-bound myo-inositol in the liver, there was no evidence of fatty liver in the young rats at any age. The cerebrum and cerebellum of myo-inositol deprived rats had normal myelination and mitochondriogenesis as judged by the levels of 2',3'-cyclic nucleotide-3'-phosphohydrolase (EC 3.1.4.1) and fumarase (EC 4.2.1.2) activity, respectively.