Distinct Structural Flexibility within SARS-CoV-2 Spike Protein Reveals Potential Therapeutic Targets

Serena H. Chen; M. Todd Young; John Gounley; Christopher Stanley; Debsindhu Bhowmik

Here, we used unsupervised deep learning routines based on CVAEs to systematically compare S protein ensembles from MD simulations across lower-dimensional, latent spaces. By first comparing the S protein protomer structure of SARS-CoV-2 to those from previous human coronaviruses, we identified distinct clusters for each virus in the 3-D latent space, where representative structures from these clusters highlight their differences in domain flexibility. Next, we compared the SARS-CoV-2 S protein protomer and trimer structures, which also displayed a clear separation of clusters in the latent space. While the main distinctions between these two states arise from the general gain in structural stability as the protomer self-assembles into the trimer state, we pinpointed structural transitions in specific flexible regions of the protomer that warrant consideration as potential therapeutic targets. These regions are promising as natural targets of immune recognition, but more importantly, they are involved in S protein oligomerization, suggesting they are susceptible to therapeutic action for protein destabilization. Overall, our study provides a more complete molecular view of the SARS-CoV-2 S protein that may assist in accelerating both vaccine and drug design efforts.