Causal Inference in Genetics: Integrative Methods for Mendelian Randomization and Single-Cell CRISPR Screening

About This Grant

This research project aims to develop innovative computational methods to enhance causal inference in genetics and genomics, addressing key challenges in understanding how genetic variation drives disease mechanisms. Advances in sequencing technologies, biobanks, and multi-omics datasets have provided unprecedented opportunities to study the genetic and molecular basis of diseases. However, significant barriers remain, including pervasive biases in Mendelian Randomization (MR), technical noise in pooled CRISPR screens, and limitations in the generalizability of genomic findings across populations. Over the next five years, this program will focus on creating statistical and machine learning tools that integrate experimental data with large-scale external datasets to overcome these challenges. For MR, the research will develop new methods to address confounding arising from family and environmental structure and to improve estimation of ancestry-specific causal effects. These methods will leverage family-based data, data integration and transfer learning, and Bayesian frameworks to strengthen robustness and improve the reliability of causal estimates, particularly for complex neuropsychiatric and related health outcomes. For single-cell CRISPR screens, the program will create advanced models that combine in silico predictions from unperturbed cells with experimental data to improve the accuracy of perturbation effect estimates and optimize experimental designs. The central goal of the project is to unify population-scale genomic studies with cellular-level experimental data to identify causal pathways linking genetic variants, cellular responses, and disease outcomes. This work will address data heterogeneity and refine causal models of genetic regulation and disease progression. The project will also support training and mentorship of students and researchers in quantitative genomics, causal inference, and computational biology. By addressing these challenges, the research will contribute to fundamental advancements in genomics, improving the understanding of complex traits and informing the development of more precise and effective therapeutic strategies.