Artem Sokolov is an Instructor in Biomedical Informatics at Harvard Medical School and serves as the Director of Informatics and Modeling at the Laboratory of Systems Pharmacology (LSP). He completed his PhD in bioinformatics at Colorado State University under the supervision of Prof. Asa Ben-Hur and trained as a postdoctoral fellow with Prof. Josh Stuart at University of California, Santa Cruz. Artem has extensive expertise in machine learning and computational biology, and his research is focused on developing robust, interpretable methods that naturally integrate with biological domain knowledge.

As the Director of Informatics and Modeling at LSP, Artem leads a team of computational biologists and software engineers who model pre-clinical and translational data across several disease areas, including cancer, Alzheimer’s Disease and tuberculosis. He plays a key role in training and mentoring a diverse group of students and postdocs, and in managing the lab’s collaborations with academic and industrial groups in multiple large-scale consortium efforts.

[Full list of publications]  [GitHub]

Selected works:

  • Paired evaluation for robust estimates of performance in machine learning models [Preprint]
  • A scalable, modular image-processing pipeline for multiplexed tissue imaging [DOI]
  • Identifying candidates for drug repurposing in Alzheimer's Disease [DOI]
  • Characterization of stemness in 10,000+ tumors across 33 different tissue types [DOI]
  • Generalized elastic nets for model regularization that forces feature localization on a pre-defined graph [DOI]
  • Prophetic Granger Causality to infer gene regulatory networks [DOI]
    • The winning method in the DREAM8 competition [DOI]
  • A stem cell signature associated with aggressive prostate cancer [DOI]
  • Structured-output method for prediction of protein function [DOI]
    • A top performer in the first CAFA challenge [DOI]
  • Unbiased tournament selection for genetic algorithms and evolutionary programming [DOI]