Welcome! I am an Assistant Professor of Biomedical Informatics at the University of Pittsburgh and a practicing child and adolescent psychiatrist in the Center for Autism and Developmental Disorders. My research develops computational methods for treatment discovery, with a natural emphasis on causal discovery and inference. We have created approaches that pinpoint root causes of disease, detect subtle treatment effects in randomized trials, and overcome latent confounding in observational studies. I gravitate toward problems that appear simple but reward deep thought, often yielding solutions that seem obvious only in hindsight. Ultimately, my goal is to produce foundational insights that translate into better patient outcomes.

Recent Inventions & Discoveries

• (Discovery) Oxytocin enhances social-emotional reciprocity in autism [paper]: We uncovered subtle, robust effects of oxytocin on social-emotional reciprocity using our SCORE algorithm – addressing a long-standing challenge in the field, where conventional total scores and subscores have repeatedly failed to detect consistent benefits.
• (Invention) Causal inference with latent confounding via time-limited treatment effects [paper]: We demonstrate that the transient nature of certain treatments enables robust causal inference, even in the presence of latent confounders.
• (Invention) Allowing algorithms to learn truly predictable clinical outcomes [paper][poster]: Outcomes are often predefined and treated as fixed, limiting discovery. Our SCORE algorithm instead lets data reveal clinically interpretable outcomes that are genuinely predictable over time, uncovering novel and actionable patterns in high-risk populations that conventional approaches overlook.
• (Discovery) Bupropion and mirtazapine show consistent differential effects among antidepressants [paper]: We identified consistent, clinically meaningful differential effects of bupropion and mirtazapine using our Supervised Varimax algorithm, addressing a long-standing problem in psychiatry where randomized controlled trials have failed to distinguish these medications despite clear clinical differences.
• (Discovery) Distinct aberrant behavior profiles reveal hidden mental distress in autism [paper]: Caregivers of patients with autism often observe non-specific behaviors, but many patients cannot communicate their distress, leaving clinicians to guess at the causes. We developed a method that links behavior patterns to specific types of mental distress, enabling clinicians to understand even non-verbal patients. This reveals actionable clinical patterns unique to autism and clarifies subtle internalizing and externalizing symptoms.
• (Invention) Learning outcomes that maximally differentiate psychiatric treatments [paper]: We developed the Supervised Varimax technique to identify outcome measures that reveal large, clinically meaningful differences between psychiatric treatments, enabling precision psychiatry in settings where standard analyses detect little to no differential effect.
• (Invention) Root causal gene discovery from high-throughput perturbations [paper]: We introduced the first method to identify root causal genes – the initial expression changes driving disease – by learning causal gene order from Perturb-seq data and transferring it to bulk RNA-seq, enabling precise patient-specific discovery of disease origins.