TOPSIS based multi criteria QSPR modeling of antibiotics using graph theoretic indices

This paper focuses on the structural features of antibiotic drugs based on topological descriptors defined according to the M-polynomial framework. Degree-based indices, i.e., Zagreb, Harmonic, Forgotten, etc., are calculated and used as molecular descriptors for the purposes of quantitative structure property relationship (QSPR) modeling. For prediction accuracy measurement, the regression models of cubic and power types are utilized and the cubic type is better chosen with larger determination coefficient and lower standard error. In addition, multi criteria decision making methods, such as TOPSIS and SAW, are applied to rank the antibiotics by computed descriptors and physical properties. Objective feature importance is guaranteed by the entropy weighting scheme. This combined approach shows the power of integrating QSPR modeling, graph theoretic indices and entropy based decision analysis for antibiotic screening that provides useful information for drug discovery and optimization.