Hongyu Zhao

This paper introduces GENJI, a method to estimate genetic correlation using both individual-level and summary-level GWAS data, improving accuracy and efficiency over existing methods.

RobustCell evaluates attack-defense methods in single-cell and spatial transcriptomic data analysis, showing attacks can impair clustering and annotation, while defenses can protect model performance.

The paper develops cell-type-specific polygenic risk scores for CAD using scRNA-seq data and identifies lipid interactions specific to adipocytes and other cell types.

CONCISE improves spatial ligand-receptor interaction inference by accounting for spatial autocorrelation and measurement errors, reducing false positives and enhancing detection power in spatial transcriptomics data.

This paper introduces PertOmni, a multimodal framework that aligns transcriptomic data with text and image embeddings to improve perturbation effect prediction.

MRMU improves Mendelian randomization by accounting for measured covariates and unmeasured confounders, identifying robust risk factors for coronary artery disease and novel drug targets.

The paper compares the developmental strategies of the enteric nervous system and intrinsic cardiac nervous system to reveal how organ-specific specialization occurs through distinct molecular and cellular mechanisms.

This study compares p-value and correlation strength for selecting gene pairs in single-cell data, finding p-values more robust for co-expression network estimation and benchmarking.

INSPIRE integrates and interprets multiple spatial transcriptomics datasets, revealing tissue structures and biological processes across diverse samples and technologies.

EA-PheWAS improves gene-phenotype discovery by integrating phenotype embeddings with PheWAS, outperforming conventional methods in detecting associations.

The paper introduces CASE, a Bayesian framework for cell-type-specific and shared eQTL fine-mapping, showing improved performance over existing methods in simulations and real data.

The paper introduces UKBioBERT, a DNA language model pre-trained on UK BioBank variants, which improves gene expression prediction and enables eQTL identification when combined with sequence-to-function models.

DrugPlayGround benchmarks large language models for drug discovery tasks like drug characteristics and interactions, assessing their performance against traditional methods.

This study evaluates foundation models for single-cell data analysis across eight tasks, finding that task-specific methods may outperform them in some cases and proposing guidelines for improving FM performance.

This paper improves genetic risk prediction across diseases by developing multi-trait polygenic scores that integrate GWAS data and show better performance than single-trait scores.

This study identifies genetic loci influencing immune cell proportions and their associations with autoimmune diseases using GWAS and cWAS.

scGPD designs gene panels for spatial transcriptomics using single-cell data, achieving better coverage and classification accuracy than existing methods.

This study benchmarks 27 computational tools for transferring cell type labels from scRNA-seq to scATAC-seq data across human and mouse tissues, finding Bridge and GLUE as top performers.

spRefine denoises and imputes spatial transcriptomics data using a reference-free deep learning framework, improving data integration and revealing new aging-related biological insights.

scELMo uses Large Language Models to analyze single-cell data for clustering, batch correction, and cell-type annotation without retraining, outperforming existing methods.

This study introduces PleioSDPR, a method that improves polygenic risk prediction by jointly modeling genetic effects across correlated traits, showing significant accuracy gains in real trait applications.

VISTA improves spatial transcriptomic data by imputing missing gene expression and enables downstream analyses like ligand-receptor discovery and spatial velocity inference.

This paper identifies multi-omic pleiotropy factors for peripheral artery disease by analyzing genetic, transcriptomic, and proteomic data across PAD, T2D, and CAD.

This paper introduces MR-CARV, a two-sample Mendelian randomization framework integrating common and rare variants to assess the causal effect of HDL-C on preeclampsia risk, showing improved statistical power and precision.

GITIII infers cell-cell interactions from spatial transcriptomics data by analyzing cell state and niche correlations using a self-supervised graph transformer model.

PERADIGM identifies rare disease-associated genes by integrating phenotype embeddings and patient similarity, outperforming traditional methods in detecting candidate genes for ADPKD, Marfan syndrome, and NF1.

This study investigates Glis3's role in ADPKD, finding that its dual inactivation with Pkd1 worsens kidney disease, affecting metabolism and circadian function.

BrainBridge integrates single-cell transcriptomic data with phenotypes to model brain biology, prioritizing genes and cell types linked to Alzheimer's, aging, and sex differences.

This paper introduces EEPRS, a framework that improves polygenic risk prediction by integrating electronic health records through phenotype embeddings, showing significant performance gains over traditional methods.

The paper introduces TCPGdb, a database integrating T cell perturbation screens to identify critical regulators of T cell function and anti-tumor persistence.