Jeremiah H. Li

GeneBench-Pro: Evaluating Multistage Statistical Reasoning in Genomics, Quantitative Biology, and Translational Biomedicine

GeneBench-Pro evaluates AI agents' ability to perform multi-stage scientific analyses in genomics and biomedicine, showing improved benchmarking over previous versions with mixed model performance.

Prioritizing embryos with lower homozygosity may reduce disease risk in children of related individuals undergoing preimplantation genetic testing

This paper proposes using embryo-level autozygosity as a biomarker to prioritize embryos with lower disease risk in preimplantation genetic testing for couples with consanguinity.

GeneBench: Assessing AI Agents for Multi-Stage Inference Problems in Genomics and Quantitative Biology

GeneBench evaluates AI agents on multi-stage genomics analysis tasks, showing current models struggle with complex inferential workflows despite partial task completion.

ImputePGTA: accurate embryo genotyping and polygenic scoring from ultra-low-pass sequencing

ImputePGTA improves embryo genotyping and polygenic scoring from ultra-low-pass sequencing by jointly inferring inheritance and parental haplotype phasing.

Comparative Evaluation of Cross-Ancestry Polygenic Risk Scoring of Type 1 Diabetes in the All of Us Cohort

This study develops and evaluates HLA-ARC, a polygenic risk score for type 1 diabetes, showing improved performance over existing methods in diverse ancestry groups.

Development and validation of polygenic scores for within-family prediction of disease risks

This paper develops and validates polygenic scores for within-family disease risk prediction, showing strong performance across ancestries and demonstrating their clinical utility in embryo screening.

Genotype inference from aggregated chromatin accessibility data reveals genetic regulatory mechanisms

The paper performs genotype inference from aggregated chromatin accessibility data to identify caQTLs and their links to gene expression and complex traits.

Ancestry dynamics and trait selection in a designer cat breed

The paper analyzes ancestry and selection in Bengal cats using sequencing data to identify genetic factors underlying coat color traits.

A comparison between low-cost library preparation kits for low coverage sequencing

This study compares four low-cost library preparation kits for low coverage sequencing, finding they all perform well in imputation accuracy, with kit choice depending on infrastructure, features, and use case.

Imputation of structural variants using a multi-ancestry long-read sequencing panel enables identification of disease associations

The paper imputes structural variants using a multi-ancestry long-read sequencing panel and identifies disease associations in large-scale datasets.

Low-pass sequencing plus imputation using avidity sequencing displays comparable imputation accuracy to sequencing by synthesis while reducing duplicates

This study compares imputation accuracy between avidity sequencing and sequencing by synthesis, finding comparable performance with lower duplication rates in avidity sequencing.

The effects of reference panel perturbations on the accuracy of genotype imputation

This study evaluates how perturbations in haplotype reference panels affect genotype imputation accuracy, finding that phase and genotype errors have a greater impact than variant density.

Low-pass sequencing increases the power of GWAS and decreases measurement error of polygenic risk scores compared to genotyping arrays

This study compares low-pass sequencing and genotyping arrays for GWAS and polygenic risk scores, finding sequencing provides higher power and accuracy.

Comparing low-pass sequencing and genotyping for trait mapping in pharmacogenetics

This study compares low-pass sequencing and genotyping arrays for pharmacogenetic trait mapping, finding that low-pass sequencing at 0.4x coverage achieves comparable imputation accuracy to genotyping arrays.

Relative matching using low coverage sequencing

This study shows that low coverage sequencing can effectively match relatives with genotyping arrays, achieving similar accuracy for close relatives and strong correlation for distant ones.