Yue Li

BLocPRS improves polygenic risk score prediction by combining locus-specific weights from different PRSs, achieving higher accuracy in UK Biobank and extending to non-European populations.

scE2TM improves single-cell embedding interpretability by integrating external knowledge, achieving better clustering and revealing biological perturbation signatures across datasets.

SpatialJEPA transfers spatial context from spatial multiomics data to dissociated non-spatial data through graph-context distillation, improving alignment and recovery of spatially organized programs in mouse brain multiomics.

MiRformer predicts miRNA-mRNA interactions and localizes binding/cleavage sites using a dual-transformer framework, achieving state-of-the-art performance on benchmark datasets.

scTimeBench benchmarks single-cell time-series analysis methods across forecast accuracy, embedding coherence, and lineage fidelity using eight datasets from four species.

This paper introduces a generative AI framework to design target-specific siRNAs for RNAi-based drug design, showing improved specificity and reduced off-target effects compared to existing methods.

scConcept improves single-cell transcriptomic analysis by creating interpretable biological concepts from gene-level data, enhancing clustering and interpretability across multiple datasets.

GFETM improves scATAC-seq modeling by integrating genome foundation model embeddings, revealing cell-state specific TF activity and kidney diabetes epigenomic signatures.

InSTaPath integrates spatial transcriptomics and histopathology images using multimodal topic learning to improve spatial domain identification and uncover gene-morphology relationships.

SpaTM is a topic modeling framework for spatial transcriptomics data that supports both annotation-guided and annotation-free analysis, showing competitiveness in spatial label prediction and clustering.

This paper introduces SSLPRS, a Polygenic Risk Score method using Spike-and-Slab LASSO, showing improved variable selection and prediction accuracy over existing methods.

The paper introduces MoEPRS, a Mixture-of-Experts framework for personalized polygenic risk prediction, showing modest accuracy improvements over single-source models across diverse biobanks.

This study identifies differential chromatin architecture and risk variants in specific brain cell types associated with major depressive disorder.

ConvNTC predicts disease-related miRNA pairs and cell-related drug pairs by modeling triplet network connections with a novel convolutional neural tensor completion approach.

CellAgentChat uses agent-based modeling to analyze single-cell data and detect cell-cell interactions, offering visualization and perturbation capabilities.

This paper introduces efficient algorithms for polygenic score inference using large-scale genetic data, achieving significant improvements in speed, memory use, and cross-population transferability.

This study integrates genotype, epigenome, and phenotype data to show immune cells contribute to bipolar disorder pathogenesis, heterogeneity, and therapy through epigenomic and genetic mechanisms.

scGRIP improves gene regulatory network inference in single-cell multi-omics data by integrating prior knowledge and explainable AI techniques.

ECLARE integrates unpaired single-cell multi-omic data using multi-teacher contrastive learning, improving cell pairing accuracy and biological structure preservation in depression studies.

CellMemory interprets out-of-distribution cells using a bottlenecked Transformer, outperforming pre-trained models and enabling accurate analysis of spatially resolved single-cell data.

MiRGraph identifies microRNA-target interactions by integrating sequence and regulatory network data, outperforming existing methods in prediction accuracy and strength estimation.

scMoE learns interpretable cell-type-specific representations from heterogeneous scRNA-seq data, improving clustering and spatial deconvolution across multiple datasets.

This paper improves power to detect gene-environment interactions by adapting SharePro to account for genetic effect heterogeneity in fine-mapping.

SharePro improves genetic colocalization analysis by handling multiple causal signals, showing increased power and efficiency in simulations and real data.

This paper introduces DendroPRS, a machine learning method that improves genotype-to-phenotype prediction across diverse ancestries by incorporating phylogenetic relationships.

SparsePro improves fine-mapping of GWAS results by integrating summary statistics and functional annotations, achieving higher power and faster computation than existing methods.

This study introduces moETM, a deep learning method for analyzing single-cell multi-omic data, and shows it identifies immune signatures and COVID-19 severity biomarkers.

This study develops a machine learning tool to integrate multi-omic data and identify biomarkers for antidepressant response, finding dynamic immune and inflammatory changes are associated with treatment success.

VIPRS improves polygenic risk prediction using variational inference, achieving high accuracy and efficiency in both simulated and real datasets.