Bağlanma afinite tahmin algoritmalarının spike: ACE2 derinmutagenez bağlanma verileri üzerinde değerlendirilmesi

COVID-19 is caused by the SARS-CoV-2 virus. The SARS-CoV-2 virus binds to the host receptor protein ACE2 via its spike protein to initiate viral entrance and viral infection. Because of their critical roles, these proteins have been widely investigated by experimental and computational approaches. Relatedly, ACE2 and spike variant binding rates were obtained by the deep mutational scanning experiments. The deep mutational scanning method provides a high throughput binding dataset and thereby facilitates benchmarking of computational tools. By taking advantage of the deep mutagenesis datasets of ACE2 and spike, we aimed to investigate the performances in predicting the impact of interfacial mutations of FoldX, EvoEF1, MutaBind2, and SSIPe as main predictors in addition to HADDOCK and UEP as naïve predictors. In addition to their overall performance analysis, we performed metric-based analysis using volume, hydrophobicity, flexibility changes upon mutations, and change of side chain properties of mutations. As a result, we found that FoldX has the highest performance with a 64% success rate. Additionally, we highlighted that none of the predictors performed well on the binding increasing mutations. As a result, we revealed that (i) the top-ranking predictor, FoldX, predicts most of the hydrophobicity increasing mutations as affinity depleting, (ii) HADDOCK tends to tag mutations with increasing volume as affinity enhancing, (iii) Conservation-based tools, MutaBind2 and SSIPe predict most of the mutations as affinity depleting. All in all, we concluded that classical binding affinity predictors are not yet sufficient to predict binding affinity changes across the interface of the host-pathogen protein system SARS-CoV-2 spike -ACE2.