Research

Birth Order and Disease Risk Across the Phenome

with Steven Kushner and Andrey Rzhetsky

Older studies of birth order typically rely on a few thousand families and test a handful of outcomes. I ran a phenome-wide scan on 10.3 million individuals from 5.1 million two-child MarketScan families, testing 569 diseases with two complementary designs: a between-family matched cohort and a within-family sibling comparison. 150 diseases show Bonferroni-significant birth-order associations, with later-borns at elevated risk for most of them (including several psychiatric, metabolic, and immunological conditions). The within-family design rules out most family-level confounders, which is the part I believe matters most methodologically.

Accelerometry for Frailty Prediction

with Yanan Long, Megan Huisingh-Scheetz, and Andrey Rzhetsky

I worked with roughly a week of free-living hip and wrist accelerometer data from older adults enrolled in a longitudinal aging study, and trained a set of models to classify frailty status at baseline and to predict 12-month frailty decline. The free-living signal carries more information than I initially expected, and a small set of engineered features (activity fragmentation, intensity bout structure) accounts for most of it. The clinical point of the project is whether a wearable stream can substitute for a clinic-based frailty phenotype, which matters for large epidemiological studies where in-person assessment is infeasible.

Cross-Disease Atlas of Human Skin Immune Cells

with Bryan Bryson and Robert Modlin

I have been assembling a harmonized single-cell atlas of human skin across 22 diseases and roughly 341,000 immune cells, split into a myeloid compartment of 16 clusters and a T/NK compartment of 21 clusters. The aim is to identify cell states that are genuinely disease-specific rather than disease-biased (these are very different things, and I have learned the hard way how easy it is to confuse them). Current findings include cancer-specific expansion of intermediate monocytes, cross-disease expansion of CXCL13+ T peripheral helper cells, and a shared Treg signature across autoimmune skin conditions. Manuscript in preparation.

Myeloid Signaling in Tuberculosis Granulomas

with Joshua Peters, Bryan Bryson, and collaborators

I contributed to a study of myeloid cell signaling in non-human primate tuberculosis granulomas, where we reconstructed conserved myeloid states across a time course of infection and associated them with IFN-γ and TGF-β signaling. My work was on applying CellTypist-based annotation to the granuloma compartments and on downstream comparative analysis. Authorship was offered after the preprint was posted.