Computational Genomics and Pangenomics
IBD detection, PBWT/GBWT/RLBWT indexing, pangenome graph analysis, local ancestry, and million-sample haplotype methods.
Zhi Lab · Yale Biomedical Informatics and Data Science
Computational Genomics and Biomedical AI
We build scalable algorithms and foundation models for population-scale genomics, electronic health records, and biomedical imaging.
Research Objective
Transform high-dimensional biomedical data into reusable infrastructure for precision medicine.
The lab develops statistical, algorithmic, and deep learning methods to connect genetic variation, clinical trajectories, and image-derived phenotypes with disease mechanisms and patient outcomes.
Research Programs
Clinical EHR Foundation Models
Med-BERT, CovRNN, multi-task fine-tuning, clinical trajectory modeling, benchmarking, and deployment-oriented validation across large-scale health systems.
AI-Powered Imaging Genetics
UDIP, image-derived phenotype GWAS, neuroimaging biomarkers, retinal imaging genetics, and multi-omics integration.
News
Latest updates
Bioinformatics Advances
TorchGWAS paper accepted
"TorchGWAS 1.0: GPU-accelerated GWAS at scale" has been accepted in its current form for publication in Bioinformatics Advances. Congratulations to Xingzhong and Ziqian!
TorchGWAS on GitHub
Communications Biology
TRACE paper accepted in principle
Saiful's paper "Learning dynamics of unsupervised deep learning reveal epoch-specific genetic architectures of brain morphology" has been accepted in principle at Communications Biology. Congratulations to Saiful!
Communications Biology
UDIP review paper accepted
Khush's invited review "A Review of Self-Supervised Imaging Phenotypes for Genome-Wide Association Studies" has been accepted for publication in Communications Biology after three rounds of review. It is a great learning resource for new team members. Khush was first author of the original UDIP paper and is now an Assistant Professor at Methodist Hospital.