Microtubule-targeting agents often have limitations to the development of resistance. Colchicine binding site (CBS) agents have several advantages compared with other tubulin inhibitors. Numerous medications in this class are less susceptible to multidrug resistance that restricts the viability of different inhibitors. In the present study, molecules that bind to the CBS of tubulin are collected from PubMed literature against the A549 cancer cell line. Regression models were established between the descriptor and IC50 value of all the compounds present in the training set based on significant molecular fingerprints using multiple linear regression (MLR). Fifteen most significant descriptors selected include Burden modified eigenvalue descriptors, PaDEL-weighted path descriptor, autocorrelation descriptor, topological distance matrix descriptor, MLFER descriptor, Barysz matrix descriptor, chi path cluster descriptor, and validated using internal and external validation parameters. The selected MLR-GA model has R2adjusted = 0.7895, Q2 CV = 0.76577, R2 pred = 0.7419, and R2 tes = 0.77373. An applicability domain is also defined so that it defines the chemical space that the model can predict. The above details suggest a good predictive model for CBS inhibitors that can predict the IC50 value of the unknown chemical compound.
Corresponding Author: Krishna kumar Ojha