Paul Kirk

TGF-β- and IL-1-dependent fibroblast states differentially shape pancreatic cancer microenvironments and metastasis

This study shows that TGF-β-dependent myCAFs promote pancreatic cancer metastasis, while IL-1-dependent iCAFs do not, through distinct signaling mechanisms.

scFLAME: a unified generative model for interpretable clustering, hierarchical structure discovery and marker-gene identification in single-cell RNA-seq data

scFLAME is a generative model that unifies clustering, marker-gene identification, and hierarchical structure discovery in scRNA-seq data, achieving state-of-the-art accuracy and scalability.

Multi-stromal organoid co-cultures model pancreatic cancer and pancreatitis epithelial cell-fibroblast heterogeneity

This study compares fibroblast heterogeneity in pancreatic cancer and pancreatitis using organoid co-cultures and identifies distinct stromal reprogramming in each condition.

Semi-supervised Bayesian integration of multiple spatial proteomics datasets

This paper introduces a semi-supervised Bayesian method to integrate spatial proteomics data with other omic datasets to improve protein subcellular localization inference.

Bayesian clustering with uncertain data

This paper introduces DPMUnc, a Bayesian clustering method that incorporates data uncertainty, and applies it to GWAS and gene expression data to identify disease subgroups.

Consensus clustering for Bayesian mixture models

The paper introduces a consensus clustering method for Bayesian mixture models, showing it can uncover target clustering structures and handle large datasets efficiently.

Transcriptional, epigenetic and metabolic signatures in cardiometabolic syndrome defined by extreme phenotypes.

The paper identifies molecular and metabolic signatures of cardiometabolic syndrome using extreme phenotype groups and shows changes after bariatric surgery.

Noise-augmented directional clustering of genetic association data identifies distinct mechanisms underlying obesity

The paper clusters genetic variants associated with body mass index to identify distinct biological mechanisms underlying obesity and its impact on coronary heart disease.

A semi-supervised Bayesian mixture modelling approach for joint batch correction and classification

This paper introduces a semi-supervised Bayesian mixture model for joint batch correction and classification, showing improved performance on low-dimensional data compared to existing methods.

A Bayesian semi-parametric model for thermal proteome profiling

This paper introduces a Bayesian semi-parametric model to analyze thermal proteome profiling data, improving sensitivity and identifying new drug-protein associations.

A fast and efficient colocalization algorithm for identifying shared genetic risk factors across multiple traits

This paper introduces HyPrColoc, a Bayesian algorithm for multi-trait colocalization using GWAS summary statistics, identifying 43 regions where coronary heart disease colocalizes with other traits.

Inferring differential subcellular localisation in comparative spatial proteomics using BANDLE

The paper introduces BANDLE, a Bayesian method to infer differential subcellular localisation from spatial proteomics data, showing improved accuracy and novel biological insights.

A semi-supervised Bayesian approach for simultaneous protein sub-cellular localisation assignment and novelty detection

This paper introduces a semi-supervised Bayesian method for protein sub-cellular localisation and novelty detection, identifying new compartments in proteomic data.

MR-Clust: Clustering of genetic variants in Mendelian randomization with similar causal estimates

MR-Clust clusters genetic variants with similar causal estimates in Mendelian randomization to identify distinct causal mechanisms.

BaseQTL: a Bayesian method to detect eQTLs from RNA-seq data with or without genotypes

BaseQTL detects eQTLs from RNA-seq data with or without genotypes, showing consistent results and identifying a psoriasis-specific effect for GSTP1.

Informed dimension reduction of clinically-related genome-wide association summary data characterises cross-trait axes of genetic risk

The paper introduces a method to reduce dimensions of GWAS summary data for related traits, identifying genetic risk patterns and enabling functional characterisation of components.

GPseudoClust: deconvolution of shared pseudo-profiles at single-cell resolution

GPseudoClust jointly infers pseudotime ordering and gene clusters while quantifying uncertainty, showing effectiveness across simulated and experimental datasets.

A Bayesian Mixture Modelling Approach For Spatial Proteomics

This paper introduces a Bayesian mixture model for spatial proteomics to probabilistically assign proteins to sub-cellular locations and quantify uncertainty in classification.