Structure-based prediction of SARS-CoV-2 variant properties using machine learning on mutational neighborhoods

Area of Science:

• Virology
• Structural Biology
• Computational Protein Science

Background:

• The SARS-CoV-2 virus, responsible for the COVID-19 pandemic, continuously evolves, with spike protein receptor-binding domain (RBD) variants posing significant public health challenges.
• Understanding the structural and functional impact of these variants is crucial for effective pandemic preparedness and response.

Purpose of the Study:

• To create a structure-enriched resource integrating theoretical and empirical data for SARS-CoV-2 spike RBD variants.
• To facilitate structure-function analysis and variant modeling for improved virology and protein science research.

Main Methods:

• Integration of large-scale in silico structure predictions (AlphaFold2, ESMFold) with empirical biophysical measurements.
• Inclusion of 3,705 single-point Wuhan-Hu-1 RBD variants and 100 Omicron BA.1/BA.2 variants.
• Annotation with structural descriptors (RMSD, TM-score, plDDT, etc.) linked to ACE2 binding and expression data from deep mutational scanning.

Main Results:

• A dataset combining computational and experimental data for a substantial number of SARS-CoV-2 spike RBD variants.
• Detailed structural and biophysical annotations correlating variant properties with functional impacts, such as ACE2 binding.
• The dataset is provided as a FAIR Data Package for accessibility and reuse.

Conclusions:

• The developed resource supports in-depth structure-function analysis of SARS-CoV-2 variants.
• This integrated dataset is valuable for variant modeling, virology research, structural biology, and computational protein science.
• The resource contributes to pandemic preparedness by enabling a better understanding of viral evolution and variant impact.