Research
Viruses are the most abundant biological entities on Earth, and almost every microbe is in an evolutionary arms race with them. We study that arms race: who infects whom, how both sides keep evolving, and the rules that shape how viral genomes are built. Our tools are field sampling, lab experiments and interpretable modern data science.
How does diversity persist?
Viruses and their hosts stay extraordinarily diverse despite constant selection. Our five-year ocean time series found viral communities that looked stable overall while their genetic variants turned over continuously, a "Red Queen" signature of coevolution seen directly in nature. We now follow virus and host populations over multiple years in the ocean, in soil and in the human gut. We ask which side adapts faster, which changes evolve repeatedly, and how a warming climate affects the pace of evolution.
What's in a genome?. Intra population diversity across three metagenomically assembled genomes (Nat. Micro. 2020)
Who infects whom?
Predicting which virus can infect which host is still one of the hardest problems in microbial ecology. We study soil bacteria and the phages that infect them, isolated close to home at the California Botanic Garden. Our questions are what molecular barriers a virus has to cross to reach its host, and how the range of hosts it can infect changes over time. We combine host-range experiments, evolutionary reconstruction and protein biochemistry, and much of this project is done by undergraduate researchers.
Replaying the tape of evolution. Biochemical characterization of ancestral state phage proteins and its relation to realized host range.
Reading viral genomes with AI
LLMs read genomes. PLMs applied to viral genomes learn to distinguish metagenomic
sequences.
In any virome, most genes have no known function. We build interpretable machine-learning models, including protein language models, that learn how viral genomes are put together. We use them to find viruses in complex samples, track genes as
they move between lineages, and generate predictions we can test at the bench.
Interdisciplinary Collaborations: data science for human health
The skills we use to analyze millions of viral genomes also apply to human health data. With clinical and public-health collaborators, we have tracked SARS-CoV-2 variants across two large California counties, made the case for microRNAs as markers of environmental and social ("extrinsic") risk in type 2 diabetes, and used the NIH All of Us cohort to study disparities, discrimination and patient–clinician communication in alcohol use disorder treatment.
Reproducible pipelines. Nextflow based pipelines are at the center of our work