Studies on the interaction between antibiotics and DNA

In this paper, 24 parameters (descriptors) that can influence the interaction of selected antibiotic compound (antibiotics) and DNA have been investigated. Principal components analysis (PCA) and hierarchical cluster analysis (HCA) were employed in order to reduce dimensionality and investigate which subset of variables could be more effective for classifying the antibiotics according to their binding mode with DNA. Results of the PCA and HCA study showed that there are 12 descriptors closely related to the interaction. Based on these descriptors, multiple linear regression (MLR) and artificial neutral network (ANN) allowed us to propose three models which can predict binding constant and binding mode. For the prediction of binding constant, the minimal relative error is only 0.17% (MLR) and 0.72% (ANN); for the prediction of binding mode (ANN), five test molecules are all consistent with experimental results.