(314f) Computational Modeling of Sars-Cov-2 Spike RBD Binding to Human ACE2 Receptor Using Molecular Simulations and Machine Learning

SARS-CoV-2 is a highly virulent pathogen which gains entry to human cells by binding with the cell surface receptor – angiotensin converting enzyme (ACE2). We used biophysics based molecular modeling to computationally contrast the binding interactions between human ACE2 and coronavirus spike protein receptor binding domain (RBD) of the SARS-CoV-1, SARS-CoV-2, and bat coronavirus RaTG13 using the Rosetta energy function. We found that the RBD of the spike protein of SARS-CoV-2 is highly optimized to achieve the strongest binding of the loft against human ACE2 (hACE2) which is consistent with its enhanced infectivity¹. On top of its high natural infectivity, SARS-CoV-2 has continued to evolve further into several novel variants with increased fitness and potential for immune evasion. Most of these variants had mutations which lead to amino acid changes in the RBD with significant effects on the receptor binding strength and dynamics. Reliably predicting the effect of amino acid changes on the ability of the RBD to interact more strongly with the hACE2 can help assess the implications for public health and the potential for spillover and adaptation into other animals. We introduced a two-step framework that first relies on recording molecular dynamics (MD) trajectories of RBD−hACE2 variants to collate binding energy terms decomposed into Coulombic, covalent, van der Waals, lipophilic, generalized Born solvation, hydrogen bonding, π−π packing, and self-contact correction terms. In the second step, we implemented a neural network to classify and quantitatively predict binding affinity changes using the decomposed energy terms as descriptors. The neural network model achieved a validation accuracy of 82.8% for classifying single–amino acid substitution variants of the RBD as worsening or improving binding affinity for hACE2 and a correlation coefficient of 0.73 between predicted and experimentally calculated changes in binding affinities². However, the need for computationally expensive MD simulations limited the use of this model for a large-scale variant surveillance. Hence, we developed a convolutional neural network (CNN) model trained on protein sequence and structural features to predict experimental RBD-hACE2 binding affinities of 8,440 variants upon single and multiple amino acid substitutions in the RBD or ACE2. This deep learning model achieved a classification accuracy of 83.28% and a Pearson correlation coefficient of 0.85 between predicted and experimentally calculated binding affinities in five-fold cross-validation studies and predicted improved binding affinity for most circulating variants. We used the trained CNN model to exhaustively screen for novel RBD variants with combinations of up to four single amino acid substitutions and suggested candidates with the highest improvements in RBD-ACE2 binding affinity for human and animal ACE2 receptors. We found that the binding affinity of RBD variants against animal ACE2s follows similar trends as those against human ACE2. White-tailed deer ACE2 binds to RBD almost as tightly as human ACE2 while cattle, pig, and chicken ACE2s bind weakly. The model allows testing whether adaptation of the virus for increased binding with other animals would cause concomitant increases in binding with hACE2 or decreased fitness due to adaptation to other hosts.³

References:

1. Chowdhury, Ratul, et al. "Computational biophysical characterization of the SARS-CoV-2 spike protein binding with the ACE2 receptor and implications for infectivity." Computational and structural biotechnology journal 18 (2020): 2573-2582.
2. Chen, Chen, et al. "Computational prediction of the effect of amino acid changes on the binding affinity between SARS-CoV-2 spike RBD and human ACE2." Proceedings of the National Academy of Sciences 118.42 (2021).
3. Chen, Chen et al. “A CNN model for predicting binding affinity changes between SARS-CoV-2 spike RBD variants and ACE2 homologues.” bioRxiv 2022.03.22.485413.