Research
I use genomics, metagenomics, machine learning and genome-scale metabolic models (GEMs) to understand microbial communities, antibiotic resistance, feed efficiency and methane production in livestock. Wet-lab training lets me design the experiments that generate the data, and computational training lets me extract the biology from it.
Ongoing
As a Mitacs Accelerate Postdoctoral Fellow and Research Manitoba Postdoctoral Fellow in the Translational Livestock Microbiome Research Lab (Derakhshani lab, University of Manitoba), I lead the computational work on two funded collaborations.
Feed is 60 to 70 percent of the cost of pig production, and host genomic selection for feed efficiency has plateaued because the trait is only weakly heritable. The gut microbiome carries much of the remaining variance. With the pig-breeding company Topigs Norsvin we sample animals from lines with divergent feed-efficiency phenotypes and ask which microbial functions, rather than which taxa, separate efficient from inefficient pigs.
In partnership with the Guan lab at the University of British Columbia, this NSERC Alliance project combines rumen microbiome data, machine learning and genome-scale metabolic models (GEMs) to predict methane phenotypes and identify ways to suppress rumen methanogens without harming the animal or the beneficial community. GEMs are built at two levels: individual methanogens, to define essential archaea-specific reactions, and the rumen community, to simulate cross-feeding, hydrogen transfer and methane output across methanogens and fermenting bacteria.
Focus Areas
Genome-resolved and culture-enriched metagenomics of gut, manure and digestate communities to track how antimicrobial resistance genes and mobile elements move through agricultural systems.
Genome-scale metabolic models (GEMs) of rumen methanogens at the individual level, to map methanogenesis and cofactor biosynthesis and prioritise essential archaea-specific enzymes, and at the rumen community level, to simulate how methanogens and fermenting bacteria exchange hydrogen and shape methane output. Both feed structure-based discovery of anti-methanogenic compounds.
Deep learning and tree-based models trained on high-throughput antibacterial screens and chemical–genetic data, used to rank compound libraries and infer mechanism of action before a compound reaches the bench.
Rationally designed pooled CRISPRi libraries targeting the essential genome of Burkholderia cenocepacia, screened against thousands of compounds to build chemical–genetic interaction profiles that expose drug targets and gene function.
Combining functional genomics with computational analysis to map how the bacterial cell envelope shapes antibiotic susceptibility, and to identify essential domains and genetic vulnerabilities in multidrug-resistant pathogens.
Methods & Tools