Overview of research program
The overall objective of our program is to leverage multi-omics (genomices, transcriptomics, proteomics, metabolomics) to identify new drug targets and to optimize pharmacotherapy decisions in patients. We are broadly interested in immune-mediated and inflammatory diseases (IMIDs), but currently focusing on inflammatory bowel disease and type 1 diabetes.
Our research is structured along three axes:
- Discover and validate drug targets for IMIDs
- Identify molecular predictors of drug response for IMIDs.
- Develop AI methods to prioritize treatments at the individual level.
Axis 1 - Discover and validate drug targets for IMIDs
Functional genetic variants that are associated with disease risk may help understand biological disease pathways. For instance, polymorphisms associated with drug target activity can be used to predict the long term on-target effects of drugs Holmes MV, NEJM (2019). Large-scale analyses have shown that drugs that are genetically-supported have increased odds of progressing through the drug development pipeline, illustrating how human genetics may be used to reduced attrition rates in drug development Minikel EV, et al. Nature (2024).
We will use high-throughput multi-omics and new machine learning algorithms to operationalize this approach towards identifying and validating new drug targets for IMIDs.
Methods: TWAS, Mendelian randomization, multi-omics
Axis 2 - Identify molecular predictors of drug response for IMIDs
PREMIIPED is an ongoing study of pediatric inflammatory bowel disease patients with multi-omics data. Measurements include treatment naive and post-treatment proteomics and transcriptomics which can be used to identify molecular biomarkers of drug response.
As a complementary approach leveraging genomic variants, we are developing partitioned polygenic risk scores to capture distinct biological mechanisms driving the genetic inflammatory bowel disease risk.
Methods: high dimensional statistics, predictive models, polygenic risk scores
Axis 3 - Develop AI methods to prioritize treatments at the individual level
Our research involves the development of new machine learning and AI methods to support our objectives. This includes the development specialized algorithms for multi-omics data fusion and causal inference. We are also interested in leveraging agentic AI solutions to identify new potential treatments at the individual level. Agentic AI autonomously leverages access to specialized bioinformatics tools and databases to make verifiable and biologically-informed treatment recommendations.
Methods: contrastive learning, agentic AI, bandits, causal inference, neural networks