Research projects

The first genome-wide look at ribonucleotide embedment in human cells: CD4+ T cells from patient blood, a stem cell line (hESC-H9), and HEK293T with and without RNase H2.

Embedded rNMPs are enriched near CpG islands and track with expression and methylation at transcription start sites. Knocking out RNase H2 introduces a template versus non-template strand bias absent in wild-type cells, consistent with rG being removed from the non-template strand, possibly by Top1.

Two human AGS orthologs (RNaseH2A-G37S and RNaseH2C-R69W) built in yeast as rnh201-G42S and rnh203-K46W.

The RNaseH2A mutant cripples enzyme activity: ribonucleotides pile up, concentrated on the leading strand of replication origins. The RNase H2C mutant showed no significant increase at all, but the sites moved. Structure shifting specificity, not throughput. A count-based assay would have called it normal.

RNASEH2A expression correlated with proliferation and cell-cycle markers across 1,000+ CCLE cell lines and 10,000+ TCGA patient samples, spanning 35+ cancer types.

Published in MDPI's Biology, and awarded a US National Science Foundation Conference Award for the poster at RNA 2021.

IGen predicts a polygenic risk score for susceptibility to 20+ infections. Built during a global pandemic, when that question suddenly mattered to everyone.

Input is a consumer sequencing result: 23andMe or Ancestry. I led and designed the computational pipeline behind it.

Differentially methylated regions in smokers versus non-smokers, found with Gaussian kernel smoothing across 900+ HIV patients in Yale's VACS cohort. The findings replicated other groups' results, and methylation varied by race.

For survival prediction against the VACS Index, support vector machines beat k-nearest neighbours. More lenient thresholding would surface more features, and likely a usable AUC.