Pattern recognition methods applied to flavonoids tested against SARS-CoV-2 main protease

Abstract

The COVID-19 pandemic, which severely affected the world at the beginning of the decade, resulted in more than half a million deaths in Brazil alone and was officially declared over as a public health emergency by the World Health Organization (WHO) in May 2023. Nevertheless, research efforts aimed at the discovery, development, and optimization of drugs to combat and/or mitigate the long-term effects of the disease have continued. In this work, a set of 14 flavonoid compounds, previously indicated by computational studies as potential inhibitors of SARS-CoV-2, was evaluated with the aim of classifying them into activity groups (low and moderate/high) according to their respective IC₅₀ values. The training set comprised baicalein, baicalin, wogonin, and wogonoside reported by Liu et al., along with ten baicalein analogs obtained from commercial flavonoid suppliers, totaling fourteen compounds. The test set consisted of six flavonoids extracted from a study conducted by Su et al., which evaluated nineteen compounds against the SARS-CoV-2 main protease (Mpro). A total of 32 molecular descriptors were calculated, of which 13 were selected, autoscaled, and used for the application of Hierarchical Cluster Analysis (HCA) and Principal Component Analysis (PCA). The selected descriptors included charge on carbon 6 (C6), frontier electron density at atom 7 (def7), frontier orbital density at atoms 6 and 7 (dof6 and dof7), energies of the highest occupied molecular orbital (HOMO), the lowest unoccupied molecular orbital (LUMO), and HOMO–1, molecular hardness, dipole moment, surface area, molecular volume, refractivity, and polarizabilityThe results showed that, compounds with high and moderate activity are predominantly described by electronic variables, mainly associated with positive loadings on the first principal component (PC1), whereas low-activity compounds are characterized by steric descriptors with negative loadings on PC1, related to molecular size and shape. HCA results demonstrated the discrimination of the compounds into two clusters—moderate/high activity (A) and low activity (B)—with some misclassifications, including low-activity compounds (M12NS, M10NS, and M3NS) grouped in cluster A and the high-activity compound M6NS not assigned to either group. Based on these findings, PCA proved to be more suitable for discriminating the studied compounds and can be considered a useful tool for the selection of flavonoids with inhibitory activity against the SARS-CoV-2 3CLpro protease.