University of Maryland School of Public Health

MATRIX Lab

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

We develop interpretable, causal, and AI-enabled methods for biomedical research, environmental health and beyond, building them into open-source software that other scientists can actually use.

From complex data to actionable biomedical insight.

What we work on

Methods built for messy biological data

High-dimensional omics data are compositional, sparse, confounded, and observational. We build statistical and machine-learning methods that respect those realities, and we implement them on real studies.

Compositional data analysis

Bias-corrected differential abundance for microbiome and other relative-abundance data, with rigorous control of false discoveries.

Multi-omics integration

Joint modeling across metagenomics, metabolomics, proteomics, and clinical data to recover signals no single assay reveals.

Causal inference

Mediation, confounding, and treatment-effect estimation in observational omics studies, longitudinal cohorts, and clinical trials.

Explainable AI

Machine-learning models whose predictions come with an interpretable account of why, a prerequisite for biomedical adoption.

Environmental health

Linking exposures to the microbiome, the metabolome, and downstream health outcomes across the life course.

Open-source software

Every method is released with documented, maintained R or Python software, so that findings are reproducible by anyone.

4,500+
Citations
17
h-index
ANCOM-BC
Bioconductor package
6
Disease areas
News

Lab updates

  • The MATRIX Lab website goes live.

  • Two new awards begin: an MPower Early Scholars grant on late-onset asthma, and the Grand Challenges 2.0 team award The Air We Share.

  • Invited talk on Predicting Mortality in Older Men Through Multi-Omics Integration at the ICSA Applied Statistics Symposium, following the same talk at ENAR in March.

  • NIEHS funds a new R21 on the health impacts of wildfires and extreme heat among end-stage kidney disease patients, with Dr. Lin as contact PI.

  • ANCOM-BC2 published in Nature Methods, extending bias-corrected differential abundance to multiple groups, covariates, and repeated measures.

  • ANCOM-BC published in Nature Communications, now a standard method for microbiome differential abundance analysis.