Ying Ma

Across 13 single-cell foundation models and 94 datasets, PCA and established methods usually match or beat them; pretrained embeddings help mainly with scarce labels, rare cells and open-set detection.

SCIGMA integrates spatial multi-omics data across modalities and platforms using a deep learning framework with uncertainty-aware contrastive learning, outperforming existing methods in domain detection and modality preservation.

GESSO is a spatially informed gene set activity scoring method that improves accuracy, interpretability, and scalability for spatial transcriptomics data.

DRIFT improves spatial transcriptomics analysis by integrating spatial context into foundation models using diffusion on spatial graphs, outperforming specialized methods.

spaGSE improves pathway enrichment analysis in spatial transcriptomics by incorporating spatial dependence, identifying biologically relevant pathways in cancer and developmental tissues.

The paper introduces ctOTVelo, a method for inferring cell type-specific gene regulatory networks from single-cell transcriptomics data, achieving state-of-the-art performance.

STING improves spatial transcriptomics clustering by integrating gene-gene and spatial relations through nested graph neural networks, achieving better clustering performance across multiple datasets.

JADE improves multi-slice spatial transcriptomics analysis by jointly aligning slices and learning shared embeddings, outperforming existing methods on benchmark datasets.

BayeSMART improves multi-sample spatial clustering by integrating AI-reconstructed single-cell data with spatial context, outperforming existing methods in accuracy and efficiency.

The paper develops ExPRSweb, an online repository of polygenic risk scores for common exposures, showing improved prediction of chronic conditions when combined with traditional PRSs.

This study evaluates the utility of breast and prostate cancer polygenic risk scores across different ancestry groups using UK Biobank data, showing significant enrichments in cancer cases within top PRS percentiles.

The paper introduces Cancer-PRSweb, an online repository of polygenic risk scores for cancer traits, evaluated in two biobanks with predictive performance metrics and phenome-wide association studies.

This study evaluates 18 dimensionality reduction methods on 30 scRNA-seq datasets to assess their accuracy, robustness, and scalability for cell clustering and lineage reconstruction.