David Ryan Koes, Ph.D.

  • Assistant Professor
  • Computational and Systems Biology

Education & Training

  • B.S. Computer Science, Carnegie Mellon University-2001
  • M.S. Computer Science, Carnegie Mellon University-2006
  • Ph.D. Computer Science, Carnegie Mellon University-2009
  • Postdoc Fellow, Computational Biology, University of Pittsburgh-2011

Research Interest Summary

My research is to develop novel computational algorithms and build full-scale systems to support rapid and inexpensive drug discovery while simultaneously applying these methods to develop novel therapeutics.

Research Categories

Research Interests

The goal of my research is to develop novel computational methods and applications that unlock the potential of computation drug discovery to revolutionize the treatment of disease. In addition to developing new computational techniques, I deploy these techniques via easy to use online application and  apply them in prospective drug discovery exercises both to aid in the discovery of new drugs and to inform my core research.

I entered the computational drug discovery research field after completing my PhD thesis in compiler optimization. Since launching this new research direction, I have developed a number of immediately useful, innovate applications that enable interactive drug discovery: gnina, Pharmit, smina, 3Dmol.js, PocketQuery, shapedb, Pharmer, ZINCPharmer, and AnchorQuery. These technologies work together to make the “big data” of chemical space accessible to any researcher with a web browser.

Pharmit is an online resource for searching large libraries of compounds, including user provided libraries, using both molecular shape and pharmacophore search. Hits can then ranked using the energy minimization routines of smina/gnina. smina and gnina are custom forks of Autodock Vina that are specially optimized for energy minimization and scoring function development.  gnina is a framework for applying deep learning to structure based drug discovery. 3Dmol.js is a plugin-free and Java-free online molecular viewer. PocketQuery is an interactive online database that uses machine learning to identify potential starting points for the design of protein-protein interaction inhibitors. ShapeDB is the first indexing algorithm for molecular shapes and supports sub-linear search of databases of molecular shapes. Pharmer is a novel general-purpose pharmacophore search technology that is an order of magnitude faster than previous technologies and enables the online screening of the purchasable compounds of the ZINC database through the ZINCPharmer web application. Finally, AnchorQuery is a specialized pharmacophore search technology that has been successfully used to search PPI-biased libraries of readily synthesizable novel compounds for inhibitors of the p53/MDM2 interaction.

Using these tools, I have participated in a number of applied drug discovery projects. We used our AnchorQuery software in the identification of inhibitors of the p53/MDM2 interaction. As part of the Teach-Discover-Treat project, we identified novel inhibitors of the Plasmodium dihydroorotate dehydrogenase enzyme (33% hit rate using our smina software with the AutoDock Vina scoring function). Ongoing projects include structure-based design of inhibitors of the profilin-actin protein-protein interaction, of the anti-cancer SHMT enzyme, of the DUSP1 phosphatase, of CYP4F2 for HETE-20 inhibition, and of TIGIT for small-molecule immunotherapy.

Representative Publications

Sunseri J, King JE, Francoeur PG, Koes DR. Convolutional neural network scoring and minimization in the D3R 2017 community challenge. J Comput Aided Mol Des. 2018 Jul 10;PubMed PMID: 29992528.

Koes DR, Vries JK. Evaluating amber force fields using computed NMR chemical shifts. Proteins. 2017 Oct;85(10):1944-1956. PubMed PMID: 28688107; NIHMSID: NIHMS991824; PubMed Central PMCID: PMC6193454.

Ragoza M, Hochuli J, Idrobo E, Sunseri J, Koes DR. Protein-Ligand Scoring with Convolutional Neural Networks. J Chem Inf Model. 2017 Apr 24;57(4):942-957. PubMed PMID: 28368587; NIHMSID: NIHMS865537; PubMed Central PMCID: PMC5479431.

Rego N, Koes D. 3Dmoljs: molecular visualization with WebGL. Bioinformatics. 2015 Apr 15;31(8):1322-4. PubMed PMID: 25505090; PubMed Central PMCID: PMC4393526.

Hochuli J, Helbling A, Skaist T, Ragoza M, Koes DR. Visualizing convolutional neural network protein-ligand scoring. J Mol Graph Model. 2018 Sep;84:96-108. PubMed PMID: 29940506.

Sunseri J, Ragoza M, Collins J, Koes DR. A D3R prospective evaluation of machine learning for protein-ligand scoring. J Comput Aided Mol Des. 2016 Sep;30(9):761-771. PubMed PMID: 27592011; NIHMSID: NIHMS814708; PubMed Central PMCID: PMC5079830.

Koes DR, Baumgartner MP, Camacho CJ. Lessons learned in empirical scoring with smina from the CSAR 2011 benchmarking exercise. J Chem Inf Model. 2013 Aug 26;53(8):1893-904. PubMed PMID: 23379370; NIHMSID: NIHMS443682; PubMed Central PMCID: PMC3726561.

Hain E, Camacho CJ, Koes DR. Fragment oriented molecular shapes. J Mol Graph Model. 2016 May;66:143-54. PubMed PMID: 27085751; NIHMSID: NIHMS778580; PubMed Central PMCID: PMC4862882.

Koes DR, Camacho CJ. Indexing Volumetric Shapes with Matching and Packing. Knowl Inf Syst. 2015 Apr 1;43(1):157-180. PubMed PMID: 26085707; NIHMSID: NIHMS555441; PubMed Central PMCID: PMC4465823.

Koes DR, Camacho CJ. Pharmer: efficient and exact pharmacophore search. J Chem Inf Model. 2011 Jun 27;51(6):1307-14. PubMed PMID: 21604800; NIHMSID: NIHMS299327; PubMed Central PMCID: PMC3124593.

Full List of Publications