Multi-omic Analytics, Translational Research, Inference & eXplainable AI

Department of Epidemiology and Biostatistics University of Maryland School of Public Health

We develop statistical methods and open-source software for high-dimensional biomedical data, and apply them in collaborations across cancer, infectious disease, aging, maternal and child health, environmental health, and clinical trials.

• Compositional data analysis: sequencing yields relative, not absolute, abundances. We estimate and correct the sample-specific sampling fraction directly, so differential abundance comes with valid confidence intervals and controlled false discovery rates.
• Multi-omics integration: correlation and latent-structure estimators that link microbiome, metabolome, proteome, and clinical layers into one coherent model.
• Causal inference: mediation analysis, confounding adjustment, and treatment-effect estimation for high-dimensional, sparse, compositional exposures and mediators.
• Explainable AI: competitive prediction paired with interpretable structure, uncertainty quantification, and feature attributions.
• Translational collaboration: methods developed against real scientific questions.

• ANCOMBC: differential abundance analysis for microbiome data with bias correction. Also on Bioconductor.
• MetVAE: a variational autoencoder (VAE) framework for untargeted metabolomics correlation analysis.

For bugs or usage questions about our software, please open an issue on the relevant repository rather than emailing; that way the answer helps everyone else too.

Huang (Frederick) Lin, PhD — Principal Investigator Assistant Professor of Biostatistics

• Email: hlin1239@umd.edu
• Phone: 301-405-2438
• Office: 4200 Valley Drive, Suite 2242, College Park, MD 20742-2611
• Lab website: https://matrix-research-lab.github.io/
• UMD faculty page: https://sph.umd.edu/people/huang-lin
• Google Scholar: https://scholar.google.com/citations?user=ysKnF2EAAAAJ&hl=en