Artificial neural networks applied to predict the characteristics of anticancer medications using K-Banhatti indices

Resumo

Genetic alterations that cause abnormal and fast cell division give birth to cancer, which can be fatal. Anticancer medications are vital to treatment and successful drug development and discovery depend on an understanding of their physicochemical behaviour. Molecular topological indices-based Quantitative Structure-Property Relationship $(QSPR)$ modelling provides a computionally effective substitute for traditional experimental investigation. In this work, we examine how well K-Banahatti topological indices $B_1(G)$, $B_2(G)$, $^mB_1(G)$, $^mB_2(G)$ and $HB(G)$ predict the physicochemical characteristics of specific anticancer drugs, such as boiling point, enthalpy and molar refraction. The correlation strength was first examined using linear regression models, which showed statistically significant connections. In particular, a strong linear association between $^mB_1(G)$ and molar refraction was highlighted. An Artificial Neural Network $(ANN)$ regression framework was further constructed utilizing a multilayer perceptron architecture in order to capture complicated nonlinear dependencies beyond linear trends. The ANN’s advantage in managing nonlinear interaction among descriptors was confirmed by its improved prediction performance, especially for molar refraction. Comparative investigation revealed that while ANN gives better predictive accuracy and generalization potential, linear regression improves interpretability. All things considered, the combination of $QSPR$ and $ANN$ enhances our comprehension of the structure-property correlations of anticancer medications and offers a productive computational avenue to promote logical drug design on oncology.