The search for polyphenolic inhibitors of BTK (Bruton’s tyrosine kinase) for the treatment of multiple sclerosis using machine learning methods

Abstract

Objective: development of a machine learning model to predict the inhibitory activity of natural compounds, predominantly of a polyphenolic nature, against Bruton’s tyrosine kinase (BTK), with the aim of identifying potential candidates for the treatment of multiple sclerosis (MS) with a favourable safety profile. Material and methods. To construct the model, data from the ChEMBL database on BTK inhibitors (CHEMBL5251) were used, containing values for the concentration required to inhibit 50 per cent of the activity (IC 50 ). The IC 50 values were converted to the pIC 50 format. Molecular structures were represented as ECFP6 fingerprints using the RDKit library. To model the ‘structure–activity’ relationship, the Bayesian Ridge Regression method from the scikit-learn library was applied. Model quality was assessed using the coefficient of determination (R2), as well as Pearson’s and Spearman’s correlation coefficients between the experimental and predicted pIC 50 values. The resulting model was used to screen polyphenolic compounds. Results. A machine learning model was developed to predict the inhibitory activity of compounds against BTK. The Pearson and Spearman correlation coefficients between the experimental and predicted pIC 50 values were 0.8; the coefficient of determination (R2) was 0.6. In a virtual screening of 21 compounds, rutin demonstrated the highest predicted activity among the polyphenols (predicted IC 50 – 24 nM). High predicted BTK inhibitory activity was also identified for luteolin, baicalein, epicatechin and quercetin. Conclusion. Machine learning methods represent a promising tool for identifying new BTK inhibitors amongst natural compounds. The results obtained indicate the potential ability of a number of polyphenols to inhibit BTK and confirm the value of further experimental investigation of these compounds as candidates for the treatment of MS.