+ Clinical Research Scientist · Quantitative Researcher
+DePauw University · DS4CABS 2026 Summer Intern
+ + ++ Reuben Addison is a clinical research scientist and quantitative researcher working at the + intersection of study design, Bayesian statistics, and biomedical data science. His work + focuses on making the assumptions behind scientific and clinical decisions explicit, + traceable, and testable. +
++ His research spans Parkinson's disease, longitudinal clinical outcomes, cardiovascular + risk, movement science, and clinical trial design. He is an Assistant Professor of + Kinesiology and Director of the Motor Control Lab at DePauw University, holds a Ph.D. in + Motor Behavior, and completed the Computational Data Analytics Track, the data science + track, of Georgia Tech's M.S. in Analytics. +
++ During the 2026 CABS Data Science Summer Internship, Reuben developed + OpenTrial, a Bayesian clinical trial design engine under the mentorship + of Prof. Shicheng Guo. +
++ OpenTrial starts with a treatment and endpoint, gathers relevant public evidence, and + produces a traceable study-design report. It integrates sources including + ClinicalTrials.gov, PubMed, openFDA, DailyMed, Open Targets, PharmGKB, and Semantic + Scholar, then connects that evidence to power, assurance, prior-sensitivity analysis, + predictive probability of success, and simulation-based design calculations. +
++ The central rule is simple: only evidence with a quantifiable effect size and associated + standard error is allowed to update the prior. Other records remain visible for provenance, + but they do not quietly influence the mathematics. +
++ Reuben treated validation as part of the research problem, not as a final software check. + OpenTrial includes 100 offline tests and 46 independent validation checks against SciPy, + statsmodels, closed-form calculations, and separately written Monte Carlo implementations. +
++ That process exposed implementation problems that ordinary unit testing had missed, + including a Type I error calculation that appeared correct because calibration and + evaluation had reused the same simulated paths, and a heterogeneity calculation that + double-counted sampling error. Both were corrected. +
++ Reuben is continuing to work at the intersection of clinical research, Bayesian trial + design, biomedical machine learning, and reproducible evidence pipelines through his lab + at DePauw and collaborations in biopharma. +
++ He is particularly interested in problems where quantitative methods can make a clinical + or development decision more transparent, reproducible, and easier to defend. OpenTrial + remains a learning-oriented research tool and is not validated for real-world clinical use. +
+