Zhi Lab
Computational Genomics and Biomedical AI
Software
The lab releases methods and code that support reproducible work across genomic data science, EHR modeling, and biomedical representation learning.
GPU-accelerated GWAS for ultra-high dimensional phenotypes. TorchGWAS streams genotype data in bounded chunks and batches covariate-adjusted linear association tests on the GPU, with direct support for PLINK 1 BED, PLINK 2 PGEN, and BGEN input.
Random Projection-based IBD Detection for ultra-fast discovery of identity-by-descent segments in biobank-scale cohorts.
Contextualized embeddings and pretraining framework for structured electronic health records and disease prediction.
PyTorch codebase for deep learning models over longitudinal electronic health record data.
Distributed representation of genes based on co-expression patterns for downstream bioinformatics modeling.
Methods for genotype calling and phasing in whole-genome sequencing data.
Deep learning framework for modeling clinical trajectories and COVID-19-related clinical outcomes from EHR data.