Paul Marjoram
University of Southern California, CA·Mathematics & Statistics
Most gene sequences stored in public databases have no known function, and testing each one in the lab is slow and costly. Computational predictions offer a faster path, but they require statistical methods that can keep pace with the flood of new data and handle inconsistent information.
- Fields
Mathematics & Statistics
Genetics
Computer Science
- Topics
functional genomics
gene function prediction
gene annotation
- Grants
- 1 known grant, through Aug 2027
What they work on
- Develop a probabilistic evolutionary model using phylogenetic trees and experimental Gene Ontology annotations to automatically predict functions of uncharacterized genes.
- Build a hierarchical modeling framework that jointly infers gene function across families of related phylogenetic trees, borrowing strength to improve prediction accuracy at scale.
- Develop pathway analysis methods based on gene network structure, making results robust to inconsistent pathway definitions across databases.
- Develop statistical methods to assess how DNA methylation conservation varies along the genome, validated with Cancer Dependency Map essentiality and GDSC drug sensitivity data.
More about the research
This project develops statistical tools for functional genomics: predicting what genes do, finding which biological pathways matter in experiments, and studying how DNA chemical tags are conserved across the genome.
Why it matters
Accurate, automatically updated gene function predictions and more reliable pathway analyses could help researchers interpret genetic data more quickly. This could improve translation of genomic findings into knowledge about human health. Methods for DNA methylation conservation, validated against cancer gene essentiality and drug sensitivity data, could inform cancer research.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Junzhou Huang
University of Texas at Arlington, TX·Computer Science
Develop multi-modal deep learning methods, including large language models, to predict the functions of thousands of poorly annotated genes and support analysis of gene-gene synergies.
Erik Wright
Baylor College of Medicine, TX·Genetics
Develop a program that finds protein-coding genes in prokaryotic genomes, using new signatures of true genes to make it more accurate than existing gene-finding tools, including for small proteins.
Thierry Chekouo Tekougang
University of Minnesota, MN·Mathematics & Statistics
Develop Bayesian statistical learning models that capture both linear and nonlinear relationships across genomic, epigenomic, transcriptomic and proteomic measurements so several omics layers can be analyzed jointly.
The school
University of Southern California: 47 labs on PI Pages, $1.3B a year on research, 882 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.