Search

ImportancePredictors consistently associated with psychosis liability and course of illness in schizophrenia (SCZ) spectrum disorders (SSD), including the need for clozapine treatment, are lacking. Longitudinally ascertained medication use may empower studies examining associations between polygenic risk scores (PRSs) and pharmacotherapy choices.ObjectiveTo examine associations between PRS-SCZ loading and groups with different liabilities to SSD (individuals with SSD taking clozapine, individuals with SSD taking other antipsychotics, their parents and siblings, and unrelated healthy controls) and between PRS-SCZ and the likelihood of receiving a prescription of clozapine relative to other antipsychotics.Design, Setting, and ParticipantsThis genetic association study was a multicenter, observational cohort study with 6 years of follow-up. Included were individuals diagnosed with SSD who were taking clozapine or other antipsychotics, their parents and siblings, and unrelated healthy controls. Data were collected from 2004 until 2021 and analyzed between October 2021 and September 2022.ExposuresPolygenic risk scores for SCZ.Main Outcomes and MeasuresMultinomial logistic regression was used to examine possible differences between groups by computing risk ratios (RRs), ie, ratios of the probability of pertaining to a particular group divided by the probability of healthy control status. We also computed PRS-informed odd ratios (ORs) for clozapine use relative to other antipsychotics.ResultsPolygenic risk scores for SCZ were generated for 2344 participants (mean [SD] age, 36.95 years [14.38]; 994 female individuals [42.4%]) who remained after quality control screening (557 individuals with SSD taking clozapine, 350 individuals with SSD taking other antipsychotics during the 6-year follow-up, 542 parents and 574 siblings of individuals with SSD, and 321 unrelated healthy controls). All RRs were significantly different from 1; RRs were highest for individuals with SSD taking clozapine (RR, 3.24; 95% CI, 2.76-3.81; P = 2.47 × 10−46), followed by individuals with SSD taking other antipsychotics (RR, 2.30; 95% CI, 1.95-2.72; P = 3.77 × 10−22), parents (RR, 1.44; 95% CI, 1.25-1.68; P = 1.76 × 10−6), and siblings (RR, 1.40; 95% CI, 1.21-1.63; P = 8.22 × 10−6). Polygenic risk scores for SCZ were positively associated with clozapine vs other antipsychotic use (OR, 1.41; 95% CI, 1.22-1.63; P = 2.98 × 10−6), suggesting a higher likelihood of clozapine prescriptions among individuals with higher PRS-SCZ.Conclusions and RelevanceIn this study, PRS-SCZ loading differed between groups of individuals with SSD, their relatives, and unrelated healthy controls, with patients taking clozapine at the far end of PRS-SCZ loading. Additionally, PRS-SCZ was associated with a higher likelihood of clozapine prescribing. Our findings may inform early intervention and prognostic studies of the value of using PRS-SCZ to personalize antipsychotic treatment.

The brain is one of the most important and complex organs in humans. On the one hand, it dictates our behavioral outcomes, such as seeing, speaking, and walking, and, through these behaviors, it determines how we interact with and perceive the world. On the other hand, its pathological outcomes, such as the development of neurodegenerative diseases, affect the longevity and quality of millions of people’s lives. Today, the brain’s neural activities can be captured by neuroimaging modalities such as functional magnetic resonance imaging (fMRI), and its behavioral markers can be recorded using smartphone sensors. Through uncovering how patterns of these large-scale data are related to the brain’s behavioral and pathological outputs, one contributes to our understanding of the biological underpinnings of knowledge, consciousness, cognition, and behavior. In parallel, by developing targeted statistical frameworks, one begins to unveil the intrinsic spatial and temporal relationships between neural signatures, behavioral features, and the brain’s pathological outcomes, and to assess and monitor neurodegenerative diseases more effectively and accurately. This thesis consists of four inter-correlated topics aimed at building targeted and efficient biomarkers using large-scale data to inquire into human behavior and brain disorders. The four topics are arranged according to their logic flow (from relatively simple to reasonably complex) and will be present in the following order: assessing outcomes of the brain using (a) static, (b) dynamic, (c) intermediating, and (d) longitudinal biomarkers. 1. Building static biomarkers. Methodological approaches are developed to identify time-averaged biomarkers to assess brain disease status and severity. This is supported by an article on developing static biomarkers to assess the status and severity of Parkinson’s disease. 2. Building dynamic biomarkers. Methodological approaches are developed to identify and isolate dynamic markers or directed information flows in the whole brain. This is supported by an article on extracting dynamic neural information flows and using them to predict cognitive flexibility in humans. 3. Building intermediating biomarkers. In a complex regulative biological system, some markers are on the one hand associated with behavior (e.g., prodromal symptoms), and on the other hand give rise to brain disorders (e.g., psychosis). Methodological approaches are developed to identify, isolate, and quantify these intermediating biomarkers or neural mediators. This is supported by an article on brain-wide functional mediation analysis. 4. Building longitudinal biomarkers. The patterns of biomarkers fluctuate over time. Identifying biomarkers whose temporal patterns are coupled with neurodegenerative disorders longitudinally is useful for assessing the development and trajectory of the diseases. This is supported by an article on developing longitudinal biomarkers to assess multiple sclerosis over time. Taken together, the methodological explorations and neurobiological discoveries in this thesis suggest that biomarker-based predictive models are promising for advancing our understanding of the human brain and behavior, and for facilitating brain illness assessment and monitoring in an automated way.

Artemisinin-based combination therapies (ACTs) have begun to fail as first-line therapies for the treatment of Plasmodium falciparum malaria in Southeast Asia. Preventing the further spread of drug-resistant parasites is a top priority for global malaria elimination campaigns. A low-cost, field-based assay to detect slow clearing ACT compounds from patient samples could allow for improved tracking of antimalarial drug use, monitoring of treatment adherence, and evaluation of therapeutic efficacy and resistance. It could also provide sensitive detection of active components in assessments of tablet quality. This research reports the development of a low-cost, rapid, fluorescent sensor for the specific detection of piperaquine and mefloquine. In order to do this, DNA aptamers were identified that bind and differentiate between small molecule partner drugs. These aptamers were selected from a library of single-stranded DNA molecules for their selectivity and binding affinity. Following their isolation by a capture-SELEX method, a structure-switching aptamer fluorescent sensor was developed. The sensor performance was optimized for the detection of drug from crushed tablets and from human serum, resulting in two fluorescent sensing platforms. The sensors were evaluated for their sensitivity and specificity in relevant sample matrices and the serum sensor was validated against an LC-MS gold standard drug detection method. Feedback from potential assay end-users was collected to inform the assay development and it is believed that these aptamer sensors will be useful tools for monitoring antimalarial drug quality and studying antimalarial drug use.

Gait analysis is an increasingly useful biomedical tool, with applications including diagnosis, treatment planning, and disease monitoring. Most gait analysis is conducted in a clinical setting using motion capture, however its greatest impact and most pertinent uses are outside the laboratory in a free-living environment. To enable the transition from clinic to real-world, wearable technologies have been developed to provide affordable, accessible, remote and unobtrusive assessment. The most popular by far are accelerometry-based devices, however their reliability as free-living tools is questionable due to: the limitations of complex gait definitions; lack of robust experimental validation; an absence of ground truth in free-living environments; and their sensitivity to a range of parameters relating to the devices themselves and how they are implemented. It is the aim of this thesis to confront each of these issues, enabling a better understanding and measurement of complex gait. To achieve this, novel spatio-temporal gait definitions and visualisations are proposed to accommodate all forms of gait. It is also hypothesised that computer vision methods applied to video from a wearable camera can extract a range of gait features in all environments, providing less sensitivity in accuracy to complex movement than current accelerometry techniques. Experimentally, a study was devised and conducted using novel protocols to better imitate free-living gait in a motion capture laboratory; as well as utilising a wearable camera and range of accelerometers. It is shown that simplistic straight protocols are not representative of complex gait, and that all wearables suffer substantial loss of accuracy when submitted to such protocols. Furthermore, given robust definitions and a simple set up, complex gait can be accurately measured by motion capture; however, gait features exist on a spectrum that is highly sensitive to definition. Both points elucidate the need for unification in the field, and the questionability of the intra-study comparability and validity of gait measuring methods. Extracting gait features from wearable camera video yielded varying results. For example, turns were detected and measured with greater than 90% accuracy, step detection precision and stride length correlation with motion capture ground truth ranged from 0.78% to 0.98% and 0.55 to 0.80 respectively depending on protocol, whereas step width could not be accurately measured at all. However, these fared comparably to all research-grade and consumer accelerometers; especially when compared step-by-step. Of these, the foot-mounted devices were by far the most reliable, whereas smartwatches and smartphones showed expected high accuracy during standard validation protocols, but this in no way transposed to imitation free-living movements.

There is clear evidence to support Rapid-Eye-Movement (REM) sleep behaviour disorder (RBD) as an early predictor of Parkinson’s disease, dementia with Lewy bodies, and multiple system atrophy. Those diagnosed with RBD enact their dreams and therefore present an abnormal characteristic of movement during REM sleep. This unusual sleep disorder presents an opportunity to better understand neurological degeneration and in-turn develop preventative medicine. However, manual polysomnograph (PSG) analysis is mandatory when diagnosing RBD, which demands complex monitoring equipment, specially trained staff, bed availability, and clinicians to interpret abundant amounts of data. Inevitably this process is time consuming and expensive, creating a demand for automated diagnostic support tools to better utilise the time of clinicians. The first aim of this thesis was to achieve automated sleep stage classification using signal processing and machine learning techniques. A total of 159 features were calculated using Electroencephalography (EEG), Electrooculography (EOG), and Electromyography (EMG) signals. After extracting the features a Random Forest (RF) algorithm was used to classify five sleep stages. With respect to REM stage classification, this technique achieved an accuracy of 93%, sensitivity of 63%, and specificity of 97% (five stage Cohen kappa of 0.63). After sleep staging, clinicians must manually identify abnormal REM sleep behaviour. Several methods have been proposed for RBD detection that use EMG recordings to objectively quantify abnormal REM movement. In this work we further develop these proven techniques with additional features that incorporate the relationship of muscle movement between sleep stages and general sleep architecture. Using an RF classifier with established and additional features, the performance of RBD detection was shown to improve upon established metrics (achieving 94% accuracy, 94% sensitivity, and 94% specificity). This technique used in combination with automated sleep staging sustained a high performance in RBD detection, with an accuracy, sensitivity, and specificity of 90%, 86%, and 94%, respectively. Lastly, this thesis investigated a minimal set of sensors to fully-automate the detection of RBD, integrating automated sleep staging followed by RBD detection without the need for cumbersome EEG sensors. This thesis demonstrated a costeffective, practical, and simple RBD identification support tool using an EOG and EMG sensor to achieve an accuracy, sensitivity, and specificity of 86%, 82%, and 90%, respectively.