Hongmin Cai

RADF completes spatial perturbation profiles by combining reference-supported population anchors with location-dependent refinement, reducing macro E-distance by 70.6%.

CANDIES denoises and integrates spatial multi-omics data using a conditional diffusion model and contrastive learning, improving downstream analysis tasks.

This paper introduces GDOT-Net, a graph diffusion optimal transport network that aligns structural and functional brain connectivity for improved diagnosis of neuropsychiatric disorders.

PRISM improves spatial multi-omics analysis by imputing missing data and identifying spatial domains, outperforming existing methods on simulated and real datasets.

HyperSTAR identifies spatial domains across resolutions using hypergraph learning, outperforming existing methods in spatial omics analysis and revealing breast cancer spatial heterogeneity.

PtsRep, a self-supervised framework for learning protein tertiary structure representations, outperforms existing methods in predicting protein stability and GFP fluorescence brightness.

This paper introduces DeepPSC, a deep learning method that reconstructs protein backbone structures from alpha carbon traces, outperforming existing methods.

CNLLRR improves scRNA-seq analysis by enhancing robustness to noise and capturing geometric relationships for better cell clustering and marker gene prioritization.