Hanbin Lee

The paper develops a model of linkage disequilibrium evolution to show how genetic background and selection strength influence variant-level architecture across populations.

The paper derives an alternative to the beta-model for genetic architecture under stabilizing selection, linking evolutionary theory to mixed models for improved prediction.

The paper develops a method to estimate variance components and predict genetic values using ancestral recombination graphs (ARGs) and linear mixed models on large datasets.

This paper introduces branch relatedness and efficient algorithms for computing branch GRMs using tree sequences, enabling large-scale genomic analysis.

The paper introduces randPedPCA, an R package for rapid PCA of large pedigrees using sparse matrices and randomised linear algebra, demonstrating its efficiency on simulated and real datasets.

The paper compares GWAS methods in admixed populations, showing standard tests are more powerful due to allele frequency heterogeneity and identifying distinct LD patterns in admixed vs. single-continental segments of the same ancestry.

This paper identifies that pseudobulk and cell-wise methods for differential gene expression analysis lead to inflated errors in multi-subject single-cell studies and recommends method selection based on study design.

MarcoPolo identifies differentially expressed genes without clustering, improving reliability by directly learning group information from gene expression bimodality.

scIntegral improves cell-type identification in scRNA-seq data by reducing error rates and enabling efficient integration of multiple donors with batch effects.