An integrated multiscale computational framework deciphers SARS-CoV-2 resistance to sotrovimab

The emergence of resistance mutations in the SARS-CoV-2 spike (S) protein presents a challenge for monoclonal antibody treatments like sotrovimab. Understanding the structural, dynamic, and molecular features of these mutations is essential for therapeutic advancements. However, the intricate landscape of potential mutations and critical residues conferring resistance to mAbs like sotrovimab remains elusive. This study introduces an integrated framework that combines interface protein design, machine learning, hybrid quantum mechanics/molecular mechanics methodologies, all-atom and coarse-grained molecular dynamics simulations, and correlation analysis. Beginning with the interface-based design and analysis, this framework elucidates the interaction between sotrovimab and the S-protein, identifying pivotal residues and plausible resistance mutations. Machine learning algorithms then facilitate the identification of potential resistance mutations using structural-sequence-binding affinity-energetics features. The hybrid quantum mechanics/molecular mechanics approach subsequently evaluates the role of mutational residues as quantum regions, assessing their impact on stabilizing the macromolecular complex. To gain a deeper understanding of the dynamic behavior of these mutations, multiscale simulations comprising all-atom and coarse-grained molecular dynamics simulations were performed, revealing their structural, biophysical and energetic impacts. These simulations complemented the static predictions, capturing the conformational dynamics and stability of the mutants in presence of glycan in the S-protein. The accuracy of the predictions is validated by correlating identified resistance mutations with clinical-sequencing data and empirical evidence from sotrovimab-treated patients. Notably, two residues, E340 at the S-protein-sotrovimab interface and Y508 distal from it, and their designs, align with clinically observed resistance mutations. Furthermore, machine learning approaches predict novel S-protein sequences with enhanced/reduced affinity for sotrovimab, validated structurally using AlphaFold. This integrated framework showcases its effectiveness in identifying potential resistance mutations, corroborated with clinical insights and offering a multidimensional strategy for decoding resistance mutations in SARS-CoV-2. Its translational relevance extends to understanding resistance mechanisms and designing novel antibody therapeutics in other systems.