Photo of Justin Silverman

Justin
Silverman
MD, PhD

Assistant Professor of Information Sciences and Technology
E375 Westgate Building
University Park, PA 16802
(814) 863-8304
Additional Title(s)
Assistant Professor of Medicine
Faculty Fellow of the Institute for Computational and Data Science
Education
Duke University, M.D., 2020
Duke University, Ph.D. in Computational Biology and Bioinformatics, 2019
Johns Hopkins University, B.S. in Physics and Biophysics, 2011
Research Interests

If it has cool math and an impactful question, I am interested. Based on my combined medical and statistical training, I tend to gravitate to problems in the analysis of biomedical data; especially, genomics and electronic health data. However, my research interests are varied and include both theoretical and applied aspects of mathematics and statistics. 

Electronic Health / Personalized Medicine

  • Statistical and machine learning methods for interpreting imperfect diagnostic tests
  • Syndromic surveillance for emerging infectious diseases
  • How can we infer changes in patient care seeking behavior during a pandemic?

Genomics

  • Statistical methods for the analysis of sequence count data (e.g., 16S microbiome surveys, bulk RNA-seq, and single-cell RNA-seq)
  • Differential expression and correlation analysis when working with count compositional data. What kinds of assumptions allow these problems to be identifiable?
  • What do zeros in sequence count data represent and how do we deal with them?
  • Batch effects and PCR amplification bias

Statistics and Machine Learning

  • Efficient and accurate methods for inferring high-dimensional Bayesian models
  • Non-Gaussian, non-linear time-series analysis
  • Compositional time-series
  • Bayesian analysis of partially identified models
  • Bayesian decision theory
  • Optimal control and sequential and active learning

Other

  • Can symmetries in probabilistic models be found computationally and exploited for faster inference?
  •  When is there a closed form transformation between two families of probability distributions?
  • How can you find a transformation that maps one family of probability models into another family (assuming the two families are topologically equivalent)?
  • Gaussian process with asymmetric kernels
  • How can you identify families of probability models with identical marginal distributions.
  • Statistical Methods for Party Planning