Kulp Lab: Artificial Intelligence (AI) and Cloud Computing
Principal Investigator: John Kulp, PhD, Computational Chemistry
The goal of this division of the Kulp lab is to create better drugs by rigorous molecular design. Kulp lab research starts with a detailed hypothesis of what protein binding interactions are required of a drug for inhibition or activation. Commercial and proprietary design tools are then used to search ligand-protein or fragment-protein simulation data for poses, ranked by predicted binding affinity, which can be assembled or optimized into custom compounds that satisfy the requirements. The designed compounds are synthesized and tested in assays. With success, this establishes a predictive model to guide subsequent lead optimization. This strategy is protein-centric, complementary to ligand-centric structure-activity relationships (SAR) approaches. The Kulp lab methodology provides access to broad chemical diversity, which is crucial for solving difficult problems in lead identification and optimization. Current ongoing projects in the Kulp lab:
A. Web-based drug design tools (www.boltzmannmaps.com)
The lab aims to provide web-based drug design tools so that researchers can access rigorous molecular modeling in user friendly format. To do this, the lab has made several key scientific breakthroughs in recent years in the field of fragment-based drug design that promise to have a transformative impact on the discovery of drugs. To make these innovations broadly available, the lab has developed a prototype Cloud-hosted web application for chemical modeling and drug discovery including fragment-based design software that is unavailable in any other commercial software package. Using Amazon Web Services (AWS) cloud servers, up to 1,000 simulations can be run in parallel.
Living longer has now become a reality as the average life expectancy is higher than in any other period in history. New scientific discoveries over the past few years have shed insight into the molecular mechanisms which are fundamental in improved human life longevity. Therapeutic areas studied include diabetes, cardiovascular diseases, neurological disorders, and cancer, the major risk factors for human mortality. The anti-aging group is focused on six targets of interest and expects experimental screening to begin in 2019. To fund the drug screening, the group plans on a few campaigns to crowdfund the project.
Deep learning, a type of machine learning or artificial intelligence, is artificial neural networks, algorithms inspired by the human brain, that learn from large amounts of data. For the Kulp lab, the team is applying deep learning to a vast array of fragment-based information. The lab has, or will soon, collect data on 1,000 fragments on 1,000 therapeutically relevant proteins. That is over 1,000,000 fragment maps. Users search the fragment maps for understanding drug binding, improving drug binding, or improving physiochemical properties. The lab collects that data about how chemists use fragments to design drugs. This wealth of information from the fragment maps to the user information will be inputs for the deep learning, AI, algorithms to determine the methods that generate the safest, most effective drugs.
- Cloudsdale, I. S.; Dickson, J. K., Jr.; Barta, T. E.; Grella, B. S.; Smith, E. D.; Kulp, J. L., 3rd; Guarnieri, F.; Kulp, J. L., Jr., Design, synthesis and biological evaluation of renin inhibitors guided by simulated annealing of chemical potential simulations. Bioorganic & Medicinal Chemistry 2017, 25 (15), 3947-3963.
- Kulp, J. L., 3rd; Cloudsdale, I. S.; Kulp, J. L., Jr.; Guarnieri, F., Hot-spot identification on a broad class of proteins and RNA suggest unifying principles of molecular recognition. PLoS One 2017, 12 (8), e0183327.
- Kulp, J. L., 3rd; Kulp, J. L., Jr.; Pompliano, D. L.; Guarnieri, F., Diverse fragment clustering and water exclusion identify protein hot spots. Journal of the American Chemical Society 2011, 133 (28), 10740-3.