Matthew Greenberg

This study identifies and corrects cross-validation bias in genomic prediction methods using CVc, showing effectiveness for most methods except Lasso and ENET.

This paper shows that re-implementing a traditional linear mixed model with a modern gradient solver can outperform deep learning models in speed and accuracy for biological data.

WgLink improves whole-genome viral haplotype reconstruction accuracy and reduces memory usage by integrating local reconstructions with L0+L1 regularization.

kTWAS combines TWAS feature selection with SKAT kernel methods to improve statistical power and identify novel disease-associated genes.

PoolHapX reconstructs microbial haplotypes by integrating genetic sharing and genomic sequencing, outperforming existing tools and revealing new insights into HIV evolution.