HPC Applications Talk
ABSTRACT
At the Center for Computational Intelligence to Predict Health and Environmental Risks (CIPHER), we have established a network of well-recognized scientists dedicated to preparation for infectious diseases. I shall focus this talk on the use of high-performance computing for drug efficacy and health intelligence.
It is difficult to predict the next outbreak or pandemic of an infectious disease. However, if and when outbreaks occur, there is a need to respond faster and more efficiently. To accelerate our response, we use high-performance computing to predict the efficacy of therapeutics and improve the speed and scale of empirical drug development. We address this challenge through computational structural biology and molecular docking tools that are widely applicable across pathogens and therapeutics.
For example, we have published results soon after the emergence of "omicron" and "kraken" variants of SARS-CoV-2. We continue to work on new variants as they are discovered and sequenced. In the case of omicron (Ford et al, 2022), we predicted weak efficacy of the current vaccine-induced and therapeutic antibodies. Moreover, we did this in silico several weeks before any wet laboratory testing could be done. In terms of health intelligence, we predicted a significant spike in cases, including breakthrough infections, due to omicron. In the case of the kraken (Ford et al., 2023), we correctly predicted that redesigned vaccine-induced and therapeutic antibodies were efficacious and that new cases and breakthrough infections would be moderate.
Going forward, our goal is to develop a novel HPC platform that scales open-source software to perform in silico molecular modeling across vast libraries of candidate molecules. We aim to do these analyses with high reproducibility. Our platform supports the modeling and docking of molecular drug candidates ranging from small molecules, to peptides, to antibody therapeutics. Moreover, our approach is disease agnostic as it relies on the use of target protein structures and a set of input molecular candidates to be evaluated under physics-based models.
When coupled with high-performance computing, this work will allow us to computationally screen drug candidates and targets at a reduced cost and increased speed over laboratory work. As a result, we can generate a prioritized list of high-scoring candidate drugs for empirical study. Our approach reduces the search space of candidates to be tested and thus speeds up wet laboratory work with the goal of increasing the rate of discovery to better deliver critical tools for response to infectious diseases.