A5074427299- Genetic Associations and Epidemiology
- Artificial Intelligence in Healthcare and Education
- Pulmonary Hypertension Research and Treatments
- Advanced Causal Inference Techniques
- Gene expression and cancer classification
- Bioinformatics and Genomic Networks
- Health Systems, Economic Evaluations, Quality of Life
- Eosinophilic Disorders and Syndromes
- Chronic Disease Management Strategies
- Explainable Artificial Intelligence (XAI)
- Machine Learning in Healthcare
- Emergency and Acute Care Studies
- Genetic Mapping and Diversity in Plants and Animals
- Diabetes and associated disorders
- Statistical Methods and Inference
- Statistical Methods in Clinical Trials
- Scientific Computing and Data Management
- Celiac Disease Research and Management
- Molecular Biology Techniques and Applications
- Vasculitis and related conditions
- Health disparities and outcomes
- Health Promotion and Cardiovascular Prevention
- Adversarial Robustness in Machine Learning
- Healthcare Policy and Management
- Mast cells and histamine
Durham University
2022-2024
The Alan Turing Institute
2020-2024
Institute of Genetics and Cancer
2021-2022
University of Edinburgh
2021-2022
University of Cambridge
2015-2021
Addenbrooke's Hospital
2016-2021
Turing Institute
2020-2021
MRC Biostatistics Unit
2018-2021
Papworth Hospital
2020
Wellcome Trust
2015-2016
Abstract Eosinophilic granulomatosis with polyangiitis (EGPA) is a rare inflammatory disease of unknown cause. 30% patients have anti-neutrophil cytoplasmic antibodies (ANCA) specific for myeloperoxidase (MPO). Here, we describe genome-wide association study in 676 EGPA cases and 6809 controls, that identifies 4 EGPA-associated loci through conventional case-control analysis, additional associations conditional false discovery rate approach. Many variants are also associated asthma six...
Rationale: Recently, rare heterozygous mutations in GDF2 were identified patients with pulmonary arterial hypertension (PAH). encodes the circulating BMP (bone morphogenetic protein) type 9, which is a ligand for BMP2 receptor.Objectives: Here we determined functional impact of and characterized plasma BMP9 BMP10 levels idiopathic PAH.Methods: Missense mutant proteins expressed vitro on protein processing secretion, endothelial signaling, activity was assessed. Plasma assayed PAH variants...
Background Inflammation and dysregulated immunity are important in the development of pulmonary arterial hypertension (PAH). Compelling preclinical data supports therapeutic blockade interleukin-6 (IL-6) signalling. Methods We conducted a phase 2 open-label study intravenous tocilizumab (8 mg·kg −1 ) over 6 months patients with group 1 PAH. Co-primary end-points were safety, defined by incidence severity adverse events, change vascular resistance. Separately, mendelian randomisation was...
Genome-wide association studies (GWAS) have been successful in identifying single nucleotide polymorphisms (SNPs) associated with many traits and diseases. However, at existing sample sizes, these variants explain only part of the estimated heritability. Leverage GWAS results from related phenotypes may improve detection without need for larger datasets. The Bayesian conditional false discovery rate (cFDR) constitutes an upper bound on expected (FDR) across a set SNPs whose p values two...
Chronic thromboembolic pulmonary hypertension involves the formation and nonresolution of thrombus, dysregulated inflammation, angiogenesis, development a small-vessel vasculopathy.
Abstract High‐dimensional hypothesis testing is ubiquitous in the biomedical sciences, and informative covariates may be employed to improve power. The conditional false discovery rate (cFDR) a widely used approach suited setting where covariate set of p‐values for equivalent hypotheses second trait. Although related Benjamini–Hochberg procedure, it does not permit any easy control type‐1 error existing methods are over‐conservative. We propose new method based on identifying mappings from...
Hallucinations are a common feature of psychosis, yet access to effective psychological treatment is limited. The Managing Unusual Sensory Experiences for First-Episode-Psychosis (MUSE-FEP) trial aimed establish the feasibility and acceptability brief, hallucination-specific, digitally provided treatment, delivered by non-specialist workforce people with psychosis. MUSE uses psychoeducation about causal mechanisms hallucinations tailored interventions help person understand manage their...
Background Risk prediction tools are routinely utilised in cardiothoracic surgery but have not been developed for pulmonary endarterectomy (PEA). There is no data on whether patients undergoing PEA may benefit from a tailored risk modelling approach. We develop and validate clinically-usable tool to predict 90-day mortality (90 DM) with the secondary aim of informing factors that influence five-year (5 YM) improvement patient-reported outcomes (PROchange) using common clinical assessment...
Emergency admissions (EA), where a patient requires urgent in-hospital care, are major challenge for healthcare systems. The development of risk prediction models can partly alleviate this problem by supporting primary care interventions and public health planning. Here, we introduce SPARRAv4, predictive score EA that will be deployed nationwide in Scotland. SPARRAv4 was derived using supervised unsupervised machine-learning methods applied to routinely collected electronic records from...
Abstract Eosinophilic granulomatosis with polyangiitis (EGPA: formerly Churg-Strauss syndrome) is a rare inflammatory disease of unknown cause. 30% patients have anti-neutrophil cytoplasm antibodies (ANCA) specific for myeloperoxidase (MPO). We performed genome-wide association study (GWAS) EGPA, testing 7.5 million genetic variants in 684 cases and 6,838 controls. Case-control analyses were EGPA as whole, stratified by ANCA. To increase power, we used conditional false discovery rate method...
Abstract Genome-wide association studies (GWAS) have been successful in identifying single nucleotide polymorphisms (SNPs) associated with many traits and diseases. However, at existing sample sizes, these variants explain only part of the estimated heritability [1]. Leverage GWAS results from related phenotypes may improve detection without need for larger datasets [2]. The Bayesian conditional false discovery rate (cFDR) [3] constitutes an upper bound on expected (FDR) across a set SNPs...
Abstract High-dimensional hypothesis testing is ubiquitous in the biomedical sciences, and informative covariates may be employed to improve power. The conditional false discovery rate (cFDR) widely-used approach suited setting where covariate a set of p-values for equivalent hypotheses second trait. Although related Benjamini-Hochberg procedure, it does not permit any easy control type-1 error rate, existing methods are over-conservative. We propose new method based on identifying mappings...
Abstract Emergency admissions (EA), where a patient requires urgent in-hospital care, are major challenge for healthcare systems. The development of risk prediction models can partly alleviate this problem by supporting primary care interventions and public health planning. Here, we introduce SPARRA v 4, predictive score EA that will be deployed nationwide in Scotland. 4 was derived using supervised unsupervised machine-learning methods applied to routinely collected electronic records from...
TREs are widely, and increasingly used to support statistical analysis of sensitive data across a range sectors (e.g., health, police, tax education) as they enable secure transparent research whilst protecting confidentiality. There is an increasing desire from academia industry train AI models in TREs. The field developing quickly with applications including spotting human errors, streamlining processes, task automation decision support. These complex require more information describe...
Machine learning is increasingly being used to generate prediction models for use in a number of real-world settings, from credit risk assessment clinical decision support. Recent discussions have highlighted potential problems the updating predictive score binary outcome when an existing forms part standard workflow, driving interventions. In this setting, induces additional causative pathway which leads miscalibration original replaced. We propose general causal framework describe and...
Abstract The Scottish Patients at Risk of Re-Admission and Admission (SPARRA) score predicts individual risk emergency hospital admission for approximately 80% the population. It was developed using routinely collected electronic health records, is used by primary care practitioners to inform anticipatory care, particularly individuals with high healthcare needs. We comprehensively assess SPARRA across population subgroups defined age, sex, ethnicity, socioeconomic deprivation, geographic...
Clinical prediction models are statistical or machine learning used to quantify the risk of a certain health outcome using patient data. These can then inform potential interventions on patients, causing an effect called performative prediction: predictions which influence they were trying predict, leading underestimation in some patients if model is updated this One suggested resolution use hold-out sets, set do not receive derived scores, such that be safely retrained. We present overview...
Prediction models frequently face the challenge of concept drift, in which underlying data distribution changes over time, weakening performance. Examples can include predict loan default, or those used healthcare contexts. Typical management strategies involve regular model updates triggered by drift detection. However, these simple policies do not necessarily balance cost updating with improved classifier We present AMUSE (Adaptive Model Updating using a Simulated Environment), novel...
The Scottish Patients at Risk of Re-Admission and Admission (SPARRA) score predicts individual risk emergency hospital admission for approximately 80% the population. It was developed using routinely collected electronic health records, is used by primary care practitioners to inform anticipatory care, particularly individuals with high healthcare needs. We comprehensively assess SPARRA across population subgroups defined age, sex, ethnicity, socioeconomic deprivation, geographic location....
Abstract Genotype-environment interaction (G×E) studies typically focus on variants with previously known marginal associations. While such two-step filtering greatly reduces the multiple testing burden, it can miss loci pronounced G×E effects, which tend to have weaker To test for effects a genome-wide scale whilst leveraging information from associations in flexible manner, we combine conditional false discovery rate results obtained StructLMM. After validating our approach, applied this...