Random forest algorithm reveals novel sites that enhanced binding affinity of human-type sialic acid receptor of H9N2 avian influenza virus

The H9N2 avian influenza virus has become the predominant subtypes in chickens in Asia for several decades, and sporadic human infections are reported. Especially, some H9N2 viruses display preference for binding to human-type alpha 2, 6 glycan receptors. This emphasizes its potential threat to the public health. However, knowledge about their molecular basis of the switch of receptor preference is still limited. Here, we propose a novel strategy for predicting amino acid sites associated with glycan-receptor binding preference based on random forest algorithm, and 12 sites, N158D, N158S, A160N, A160D, A160T, T163I, T163V, V190T, V190A, D193N, D193G, and N231D, are predicted to be tend to prefer human-type alpha 2, 6 glycan receptors. Receptor binding assays proved that except the V190T site, as expected other sites displayed an affinity for preference binding to alpha -2,6SA. 158, 160, and 163 residues identified in this study did not place in 130-loop, the 190-helix and the 220-loop, which form the edges of the receptor-binding site. Thus other regions should also influence the receptor binding preference of H9N2. Especially, A160T has the greatest improvement of the preference for alpha 2,6 receptor and enhances the thermal stability. This mutation increased its replication in chicken lungs in early-stage infection, and caused significant upregulation of the immune genes and increased mortality rate in mice, this reveals its pandemic potential. In summary, we developed a novel method and revealed 11 amino acid sites that associated with receptor binding preference with high accuracy. Our finding provides novel insights in understanding the genetic basis of receptor preference of H9N2.