Michael P. Epstein

This study uses multimodal transcriptomic data to enhance Alzheimer's disease risk gene discovery, identifying 161 risk genes with proteomic support and an interconnected network of hub genes.

This study identifies de novo structural variants in individuals with orofacial clefts and finds they contribute to genetic risk, with some variants disrupting known risk genes and suggesting new candidate genes.

STACCato identifies condition-related cell-cell communication events using supervised tensor analysis, improving inference over existing methods in scRNA-seq datasets.

HiCAT improves spatial omics annotation by combining unsupervised clustering with atlas guidance, revealing molecularly defined regions and enhancing downstream analysis.

This study uses a multi-tissue transcriptome-wide association study to identify 29 risk genes for ADHD, including six previously unlinked genes.

This paper introduces KDCA, a framework for differential co-expression analysis of general risk factors, showing improved power and reduced bias compared to existing methods.

The paper identifies rare variants in PRKCI as a novel cause of Van der Woude syndrome and other peridermopathy features through clinical and functional analyses.

MorphLink identifies disease-related morphological-molecular interplays in multi-modal spatial omics data by linking interpretable morphological features with molecular measurements.

CADET improves transcriptome-wide association studies in admixed populations by integrating local ancestry and eQTL data to enhance gene expression prediction and association testing.

The paper identifies protein-altering de novo variants in CP probands, finding higher enrichment in syndromic cases and distinct biological enrichments in single-cell RNA sequencing datasets.

The paper identifies sex-specific genetic risks for cleft palate through genome-wide gene-by-sex interaction analysis in a trio-based cohort.

SCAMPI detects SNP interactions across multiple traits by analyzing trait correlation patterns, improving power over univariate methods in lipid-related traits from UK Biobank data.

This paper introduces LIT, a method to detect latent genetic interactions using multiple traits, showing improved power compared to univariate methods in obesity-related traits.

This study performs cis- and trans-eQTL TWAS on breast and ovarian cancer data to identify 101 and 8 risk-associated genes, respectively, with several driven by trans-eQTLs.

This study performs trio-based GWAS to identify genetic risk factors for cleft palate subtypes, finding a genome-wide significant locus and subtype-specific associations.

The paper uses whole-exome sequencing to identify rare variants in multiplex families with orofacial clefts, finding variants in seven families but no clear evidence that subclinical phenotypes clarify inheritance patterns.

SR-TWAS improves TWAS power by combining multiple reference panels and regression methods through ensemble learning, identifying risk genes for Alzheimer's and Parkinson's diseases.

The paper estimates the diagnostic yield of sequencing 418 genes in OFC cases and controls, finding a higher prevalence of likely pathogenic variants in cases.

POIROT detects parent-of-origin effects in GWAS by comparing phenotypic covariance matrices across genotypes, showing improved power and robustness to non-normality.

This study identifies rare variants in SEC24D associated with orofacial cleft subtypes and shows they affect transcription factor binding sites.

The paper introduces a pruning and thresholding method for methylation risk scores to improve prediction across multi-ancestry populations, showing better performance than simple thresholding.

OTTERS is a TWAS framework that uses summary-level eQTL reference data to estimate eQTL weights and perform omnibus TWAS, showing improved power and applicability.

The paper introduces PWAS-O, an omnibus proteome-wide association study method that identifies 43 risk genes for Alzheimer's disease dementia, including 5 novel ones.

The paper identifies OTUD7A loss-of-function mutations in schizophrenia-associated 15q13.3 deletions and shows they impair synapse development and function in human neurons.

The paper introduces a unified likelihood framework for mediation analysis in case-control studies to improve efficiency of indirect effect estimates by leveraging exposure-mediator relationships.

The paper identifies DNA methylation in the CLDN5 locus as associated with cognitive decline, independent of neuropathology, suggesting a role for the blood-brain barrier in cognitive trajectory.

TIGAR-V2 improves TWAS efficiency by directly reading VCF files, enabling parallel computation, and reducing computation cost by up to 90%, while identifying risk genes for breast and ovarian cancer.

The paper introduces a kernel method to dissect genetic signals in high-dimensional phenotypes, showing it controls type I error and identifies direct effects in GWAS data.

The paper identifies a genetic modifier, FAT4, associated with left-sided cleft lip and palate, suggesting genetic factors differ from epidemiological patterns.