Computational Design of Peptides to Block Binding of the SARS-CoV-2 Spike Protein to Human ACE2

Xiaoqiang Huang; Robin Pearce; Yang Zhang

However, we can also see from Figure 2E that it does not necessarily mean that higher sequence identity always ensures better designs, since the two designs with the highest sequence identity (15/31=48.4%) did not always show stronger binding than those around 35%. Thus, the results suggest that good binders showed a high similarity to the wild-type, but the similarity should not be too high in order to leave room for the designs to be improved. This is in line with the common thinking that the critical binding residues (i.e. hot spot residues) should be conserved while some other residues can be mutated to enhance binding. Note that the wild-type peptide was comprised of a helix (a. a. 22-44) and a short loop (a.a. 351-357) with a glycine linker. To ensure good binding to SARS-CoV-2 RBD, the designed peptides should be able to preserve the secondary structure of . CC-BY-NC-ND 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.03.28.013607 doi: bioRxiv preprint this motif. To check this point, we used an artificial neural network-based secondary structure predictor 28 implemented in EvoDesign to predict the secondary structure of the designed peptides; the predictor that we used here was much faster than some other state-of-the-art predictors, e.g. PSIPRED 35 and PSSpred 36 , but showed similar performance 28 . To quantify the similarity between the secondary structure of a designed peptide and that of the wild-type, we calculated the secondary structure match rate, which was defined as the ratio of the number of residues with correctly assigned secondary structure elements (i.e. helix, strand and coil) to the total number of residues (i.e. 31). As shown in Figure 2F , 892 out of the 992 designed peptides had >90% secondary structure elements predicted to be identical to that of the wild-type peptide, indicating the high accuracy of the designs, although the EvoEF2 scoring function does not include any explicit secondary structure-related energy terms 18 .