QSAR analysis of drugs using graph based degree based topological indices and regression models

Abstract

Abstract Drugs are chemical solutions that are extensively used in diagnosing, prevention and treatment of diseases. To develop the drugs, it is important to understand the correlation between the drugs structure and their physicochemical behavior. Molecular network analysis is a systematic analysis of structural features, with topological indices having an important role in the measurement of molecular architecture. Ten popular topological indices, including Atom-Bond Connectivity (ABC), Randici (RI), Geometric-Arithmetic (GA), Sum-Connectivity (SC), the first and second Zagreb indices ( $$M_1$$ and $$M_2$$ ), Schultz second index (SS), Harmonic (H), Hyper-Zagreb (HZ), and the Forgotten index were used on nine drugs, which are linolenic acid, serine, methionine, tyrosine, cystine, succinic acid, N-acetylglucosamine, glutamic Eight basic physicochemical properties were taken into account and the efficacy of the indices was examined by three types of regression straight, logarithmic and quadratic regression. The results of the analysis have shown that there are strong correlations between the chosen topological indices and the physicochemical properties which prove the usefulness of graph-theoretical descriptors in QSPR modeling. The quadratic regression technique was the most predictive of the three models used, and it was better than the linear and logarithmic models. These results indicate the high predictive potential of the topological indices especially in conjunction with non-linear modeling in the interpretation of drug structure-property correlations.