Mulin Jun Li

The paper introduces hEVO, a genomic foundation model trained on human haplotype sequences, achieving top performance in variant interpretation and functional annotation across multiple benchmarks.

The paper uses single-cell multiome data to identify QTLs regulating gene expression and chromatin accessibility during cellular senescence, revealing new causal variants and regulatory mechanisms.

The China Brain Multi-omics Atlas Project (CBMAP) creates a comprehensive molecular reference map of over 1,000 human brains to address underrepresentation in East Asian populations and advance research on aging and neuropsychiatric disorders.

GGI-RUNNER identifies rare variant gene interactions linked to complex diseases and reveals interconnected gene networks.

This study introduces a deep learning framework, FIND, to better classify genetic variants by their deleteriousness and trait-modulating effects, showing improved performance over existing methods.

This study identifies causal genes and transcriptomic factors influencing human telomere length using placental tissue data and GWAS.

The paper develops DeepAnchor, a deep learning model to characterize CTCF binding patterns and predict CTCF-anchored loops across human tissues, revealing sequence and chromatin features that influence loop formation and disease mechanisms.

The paper identifies causal regulatory elements in complex genetic loci using CRISPR perturbations and shows that many causal variants lack classical epigenetic markers.

This paper introduces a new statistical method, RUNNER, to detect rare variants associated with complex diseases by comparing observed mutation burdens to predicted baseline values, showing improved power over existing methods.

MCGA identifies schizophrenia susceptibility genes using a multi-strategy framework integrating isoform-level eQTLs, revealing genes enriched in neuronal signaling and drug targets.

This paper uses Mendelian randomization to identify causal links between coagulation factors and COVID-19 severity, finding VWF and ADAMTS13 as significant factors.

This study compares multiple Mendelian randomization methods for causal gene inference in complex phenotypes, identifies two robust methods, and applies them to psychiatric diseases to discover known and novel causal genes.

This paper develops regBase, a database of whole genome base-wise aggregations of regulatory variant predictions, and three composite models to score functional, pathogenic, and cancer driver variants, showing superior performance in fine-mapping causal variants.

WITER improves cancer-driver gene detection by accurately modeling background mutations, identifying 178 driver genes across 26 cancers with high validation rates.

The paper identifies genetic variations affecting ceRNA regulation by analyzing RNA-seq data and GWAS results, revealing new regulatory mechanisms and disease associations.