Yu Xu

Enhanced power and transferability for genetics-driven metabolomic biomarker discovery in admixed American cohorts

The paper improves metabolomic biomarker discovery in admixed American cohorts by training models on diverse data, achieving better predictive performance and more significant associations with cardiometabolic diseases.

Performance of deep-learning based approaches to improve polygenic scores

This study evaluates deep-learning methods for improving polygenic scores, finding limited benefits due to confounding effects from linkage disequilibrium.

Genome-wide analyses of variance in blood cell phenotypes provide new insights into complex trait biology and prediction

This study identifies genetic loci influencing variance in blood cell traits and shows that incorporating variance polygenic scores improves prediction accuracy.

Metabolic reaction fluxes as amplifiers and buffers of risk alleles for coronary artery disease

This study identifies metabolic reaction fluxes that interact with CAD risk alleles, revealing their role as amplifiers or buffers in disease susceptibility.

Genetic determinants of blood gene expression and splicing and their contribution to molecular phenotypes and health outcomes

This study identifies genetic determinants of blood gene expression and splicing, linking them to molecular phenotypes and health outcomes through QTL analysis and colocalization.

Deep learning for polygenic prediction: The role of heritability, interaction type and sample size

This study evaluates deep learning methods for polygenic prediction, showing that MLP benefits most from larger sample sizes and one-hot encoding improves performance across methods.

An atlas of genetic scores to predict multi-omic traits

The paper develops genetic scores for 17,227 molecular traits using multi-omic data from INTERVAL and validates them across ancestries, demonstrating their utility in pathway analysis and disease association discovery.

Genetically personalised organ-specific metabolic models in health and disease

The paper develops genetically personalised organ-specific metabolic models to identify associations between metabolic fluxes and disease risk, finding 97 fluxes linked to coronary artery disease.

The Polygenic Score Catalog: an open database for reproducibility and systematic evaluation

PGS Catalog standardizes 192 curated polygenic scores and metadata (86 traits); reapplied in UK Biobank, eight of nine colorectal cancer scores predicted disease in European-ancestry participants.

Learning polygenic scores for human blood cell traits

This study compares machine learning and univariate methods for constructing polygenic scores for 26 blood cell traits using UK Biobank and INTERVAL data, showing improved performance with learning methods.