Meta-regression models
We’ve developed a class of meta-regression models designed to integrate variant level features including constraint, pathogenicity, loss-of-function annotation, and evolutionary scores. It improves power over standard rare variant genetic analysis as is showcased in our epilepsy application.
Bayesian summary statistic framework
We’ve developed a Bayesian framework for rare variant aggregation analysis that uses a full Kronecker expansion of the prior matrices.
We’ve also developed a Bayesian clustering algorithm for rare variant effect profiles.
Penalized regression for ultrahigh-dimensional problems
We’ve developed algorithms for fitting penalized regression to ultrahigh-dimensional problems.
- snpnet
- sparse-snpnet and snpnet-2.0 with selective inference
- snpnet-Cox
- multiSnpnet Fast Multi-Phenotype Sparse Reduced Rank Regression on Genetic Data
- multiSnpnet-Cox, mrcox Efficient Group-Sparse Lasso solver for multi-response Cox model
- multiSnpnet-Cox-gpu GPU version of multiSnpnet-Cox
- multiSnpnet-Cox-tv, Coxtv Regularized Cox proportional hazard model with time-varying covariate
- regularizedMR Regularized Mendelian randomization
Causal inference methods
We’ve developed methods for causal inference from population biobank data.
Decomposition of genetic effect methods
We’ve developed methods to decompose genetic effects from large-scale human genetic studies and project them to individual risk.
Additional polygenic risk score methods
We’ve developed methods that train sex-aware polygenic risk models, and that use genetic risk from biomarkers in addition to disease risk scores.
Visualization and inference engine
We’ve developed Global Biobank Engine, software for visualizing genetic association results from population biobank data.
Artificial intelligence tools
We’ve developed artificial intelligence tools for tumor board discussions, and interpreting rare variants from whole genome sequencing data.