CDS&E: Simulation- and Data-driven Peptide Antibody Design Targeting RBD and non-RBD Epitopes of SARS-CoV-2 Spike Protein

COVID-19 has led to heavy impacts globally. Development of therapeutics and vaccines is thus of exceptional significance. Nevertheless, the design of safe and effective therapeutics and antibodies is highly challenging due to the lack of the fundamental understanding of the molecular interactions between drugs and the viral proteins, which have been consistently mutated. Designing therapeutics using the bottom-up method is a highly promising approach as we can gain more insight into the underlying molecular interactions that govern the efficacy of the drugs. However, few computation- and data-enabled efforts to search for therapeutics that reduce the host cell binding propensity of the virus could explain and predict the performance of a given drug before synthesis and testing. To address this challenge, we propose to develop a hybrid machine learning-simulation (MLSim) platform that integrates data inference and molecular simulation to enable high-throughput, high-fidelity screening, and optimization of peptide design. In addition to the wild-type SARS-CoV-2 spike protein, the MLSim platform will be employed to design peptide antibodies for recently reported spike protein mutants, an aspect barely considered in designing COVID-19 therapeutics and antibodies. We will design mixtures of peptides (cocktail peptides) targeting different epitopes simultaneously to suppress the antibody-dependent enhancement and neutralize SARS-CoV-2 spike protein mutants. The contribution of the proposed work is significant because the developed platform will 1) enable the high-throughput design of therapeutics for COVID-19 in particular and peptide-based drugs in general, and 2) help to reveal the molecular interactions between drugs and the target viral proteins.