Matthew Stephens

Torino, a reference-free framework for isoform discovery, identifies RNA processing patterns linked to disease risk by analyzing read coverage variation across species and disease states.

The paper introduces immgenT-GP, a gene-program framework that identifies 200 reproducible gene programs in mouse T cells, capturing lineage identity, activation states, and tissue location.

The paper investigates whether persistent inflammation in Crohn's disease is due to environmental triggers or intrinsic epithelial changes by comparing organoid and tissue data from patients in remission.

The paper introduces mvSuSiE, a multi-trait fine-mapping method that improves causal SNP identification by jointly analyzing multiple traits, discovering more causal SNPs and narrower credible sets compared to single-trait approaches.

SuSiE 2.0 improves genetic fine-mapping and phenotype prediction with faster performance, better calibration, and new methods like SuSiE-ash.

The paper examines T cell changes in mammary glands during pregnancy and lactation, finding an expansion of T effector cells and similarities to intestinal T cells.

The paper identifies response eQTLs associated with environmental exposures in human cells and finds they overlap with disease-associated loci not explained by steady-state eQTLs.

This paper introduces GBCD to decompose tumor transcriptional heterogeneity in scRNA-seq data, identifying a new prognostic gene expression program in pancreatic cancer.

The paper introduces an efficient algorithm for dimensionality reduction of single-cell RNA-seq data, showing it outperforms existing methods in speed and quality.

This paper introduces GoM DE, a differential expression analysis method that allows cells to have partial membership in multiple groups, improving the interpretation of structure in single-cell sequencing data.

The paper introduces a flexible empirical Bayes method for multivariate multiple regression to improve prediction of multi-tissue gene expression from genotypes.

This paper introduces a method to adjust for genetic confounders in transcriptome-wide association studies, improving reliability in detecting causal genes.

This paper extends the SuSiE model to work with summary data, showing it is competitive with existing methods in speed and accuracy for fine-mapping genetic variants.

The paper clarifies confusion in scRNA-seq analysis by distinguishing between measurement error and true gene expression, advocating for a Poisson measurement model.

The paper identifies dynamic eQTLs in cardiomyocytes under hypoxic stress, showing how genetic variation affects gene expression differently across conditions.

The paper uses GTEx data to analyze GWAS loci, showing dose-dependent RNA effects and proposing target genes for 47% of loci using combined methods.

The paper presents a large ASE resource from GTEx v8 with 431 million SNP-level and 153 million haplotype-level measurements, and extends phASER to estimate cis regulatory variant effect sizes.

The paper uses bulk RNA-seq data to estimate cell type abundances and identifies cell type interaction QTLs that provide finer resolution of tissue specificity than bulk cis-eQTLs.

The paper examines telomere length variation across 6,391 tissue samples from 952 individuals, showing tissue-specific differences, positive correlations between tissues, and associations with age and ancestry.

The paper quantifies protein levels across 32 human tissues and compares them to transcriptome data, revealing insights into post-transcriptional regulation and disease mechanisms.

SuSiE writes sparse regression coefficients as a sum of single-effect vectors; its IBSS algorithm returns a credible set per effect and outperformed existing fine-mapping methods in numerical experiments.

This paper introduces a new Mendelian randomization method, CAUSE, that accounts for correlated and uncorrelated pleiotropic effects, showing improved robustness in detecting true causal relationships.

The paper develops a method to quantify cell cycle progression in scRNA-seq data using a small number of genes, showing it can predict cell cycle phase within 14% accuracy.

This paper applies multivariate analysis to 13 GWAS datasets, identifying 433 new genetic associations that replicate in follow-up studies.

GTEx v8 maps expression and splicing QTLs in 49 tissues from 838 donors: cis-eQTLs cover 94.7% of protein-coding genes, and cross-tissue QTL sharing mirrors cell-type composition.

driverMAPS improves cancer driver gene detection by modeling positive selection and mutation patterns, identifying 159 new potential driver genes including RNA methyltransferases.

This study identifies variance QTLs (vQTLs) in human iPSCs using scRNA-seq, finding that genetic variants affect gene expression variance but with smaller phenotypic impact than mean expression QTLs.

The paper introduces a method to estimate time-varying migration rates and population sizes using genetic similarity data, applied to European populations to analyze recent demographic history.

The paper introduces a new sequence logo plot, EDLogo, that visualizes both enrichment and depletion of characters in biological sequences and provides an R package for its implementation.

The paper introduces flexible statistical methods for genomic studies with multiple conditions, improving power and effect estimates by accounting for arbitrary correlations in effect sizes.