Ira S. Hofer

This study evaluates methylation risk scores (MRS) in clinical datasets, showing they improve EHR imputation and predict phenotypes better than polygenic risk scores or EHR-based methods alone.

This paper develops a machine learning model using EMR data to predict in-hospital mortality preoperatively, achieving higher accuracy than existing clinical scores.

This paper introduces a Bayesian method to estimate cell type composition from DNA methylation without requiring a methylation reference, showing it can identify individual cell types rather than just linear combinations.