Soyeon Kim

The paper investigates reliability transfer in single-cell foundation models across biological contexts and finds it varies independently from prediction transfer.

eQTM Atlas, an R Shiny resource, curates over 11 million DNA methylation-gene expression associations from eight cohorts and 11 tissues to annotate EWAS CpGs beyond proximity.

sigNATURE improves reproducibility of ICI response biomarkers by mapping T-cell states across cohorts using reference atlases and identifiability scores.

DeepDiff-SHAP identifies subgroup-specific causal relationships in biomedical data using deep learning and SHAP, revealing clinically meaningful changes in feature interactions across patient subgroups.

DeepEXPOKE uses deep learning with knockoff features to distinguish genetic and non-genetic exposure risks in sepsis and coronary heart disease, identifying metabolites and asthma as risk factors.

Methyl-TWAS improves gene expression imputation and TWAS by using long-range DNA methylation markers, outperforming existing methods in nasal epithelium and white blood cells.

The paper identifies VNTRs near SLC9A3 associated with gene expression in CF nasal epithelia and shows their utility in explaining GWAS signals.

This paper introduces causalDeepVASE, a deep-learning method for causal inference that identifies both direct and indirect causal relationships in biological systems, outperforming existing methods in simulated and biological data.

The paper introduces scMAPA, a tool to identify cell-type-specific APA genes in scRNA-Seq data, showing improved power and interpretability in human and mouse datasets.

The paper identifies methylation-gene expression associations in nasal epithelium from Puerto Rican children with and without asthma, revealing epigenetic regulation of gene expression in atopic asthma.

This study develops geneEXPLORER, a model that predicts gene expression using long-range DNA methylation effects, showing it can accurately predict clinical phenotypes in cancer.

The paper identifies microRNAs whose target sites are modified by global APA events in breast cancer, showing their role in tumor progression and patient-specific heterogeneity.