(639e) Computational Prediction and Experimental Validation of Antibody Repurposing Via Proteinmpnn-Directed Mutagenesis

The repurposing of antibodies against existing viral threats, such as SARS-CoV, to target emerging viruses offers a promising avenue for therapeutic intervention. In this study, we utilized advanced computational techniques, specifically the Protein Message Passing Neural Network (ProteinMPNN), to identify key residues for mutagenesis to enhance the efficacy of antibodies against SARS for combating SARS-CoV-2. ProteinMPNN is a state-of-the-art tool for predicting a protein sequence that corresponds to a given protein structure. While our calculations show that it has a ~40% accuracy for correctly predicting the amino acids of antibody paratopes, we hypothesize that it can be used to guide antibody repurposing experiments.

Three neutralizing SARS-CoV antibodies, m396, 80R, and S230, with known experimental structures were computationally docked with the SARS-CoV-2 receptor binding domain. The structures of the antibodies in complex with SARS-CoV and SARS-CoV-2 were then run through a CHARMM fixed-backbone energy minimization followed by a Rosetta all-atom energy minimization to prepare them for ProteinMPNN analysis. When ProteinMPNN predicts an amino acid sequence for a protein, it provides a probability for every amino acid at each position. By comparing the probabilities at each position of the antibodies in complex with SARS-CoV versus SARS-CoV-2, it is possible to determine which positions ProteinMPNN considers most important to mutate in response to the changed antigen. The ProteinMPNN analysis progressed in an iterative manner, where in each iteration the position with the smallest probability change was fixed to its wildtype sequence. In this way, we accumulated lists of positions (10, 15, and 6, respectively) that ProteinMPNN considered most important to mutate to repurpose the antibodies to bind SARS-CoV-2 instead of SARS-CoV. This presentation will provide a detailed overview of our method and a discussion of our computational and experimental results from mutating the identified residues.