Ruowang Li

G2DBridge: A Multimodal Framework Linking Genetics to Disease through Imaging Intermediates

G2DBridge links genetic variation to disease risk by modeling imaging-derived phenotypes as intermediate traits, improving Alzheimer's Disease prediction over polygenic risk scores without new data.

A Deep Ensemble Encoder Network Method for Improved Polygenic Risk Score Prediction

This paper introduces DEEN, a deep learning method that improves polygenic risk score prediction by modeling both linear and non-linear SNP effects, showing better performance on UK Biobank and All of Us datasets.

Beyond P-values: A Multi-Metric Framework for Robust Feature Selection and Predictive Modeling

The paper introduces MIXER, a framework that combines multiple metrics for feature selection and predictive modeling, outperforming individual criteria in Alzheimer's disease analysis.

A One-Shot Lossless Algorithm for Cross-Cohort Learning in Mixed-Outcomes Analysis

This paper introduces mixWAS, a one-shot, lossless algorithm for integrating distributed EHR datasets to identify cross-cohort genetic associations across multiple outcomes.

Automated quantitative trait locus analysis (AutoQTL)

AutoQTL automates QTL analysis and GWAS by applying machine learning to detect both additive and non-additive genetic effects in genetic data.

Polygenic risk vectors (PRV) improve genetic risk stratification for cardio-metabolic diseases

This paper introduces polygenic risk vectors (PRV) to improve genetic risk stratification for cardio-metabolic diseases by combining additive and dominant genetic effects, outperforming traditional polygenic risk scores.

Evaluation of phenotyping errors on polygenic risk score predictions

This study evaluates how phenotyping errors affect polygenic risk score (PRS) accuracy using eMERGE data, showing accuracy degradation depends on phenotype and error mechanism.

Collective feature selection to identify crucial epistatic variants

The paper introduces a collective feature selection approach to identify epistatic variants by combining multiple methods, showing improved performance in detecting true positive interactions in both simulated and real BMI data.