QSAR study on the interactions between antibiotic compounds and DNA by a hybrid genetic-based support vector machine

Studies on the interactions of antibiotic compounds with DNA can provide useful suggestions and guidance for the design of new and more efficient DNA-binding drugs. A quantitative structure-activity relationship (QSAR) study of the binding modes and binding affinities of the interactions between 30 antibiotic compounds and DNA was performed. A large number of descriptors that encode hydrophobic, topological, geometrical, and electronic properties were calculated to represent the structures of the antibiotic compounds. Aiming at a system with small, multidimensional samples, we utilized the genetic algorithm-support vector machine (GA-SVM) method to develop the QSAR, which can select an optimized feature subset and optimize SVM parameters simultaneously. A binary QSAR model for predicting binding mode and conventional QSAR models for predicting binding affinity were built based on the GA-SVM approach. The descriptors selected using GA-SVM represented the overall descriptor space and can account well for the binding nature of the considered dataset. The descriptors selected using the GA-SVM method were then used for developing conventional QSAR models by the artificial neural network (ANN) approach. A comparison between the conventional QSAR models using GA-SVM with those using ANN revealed that the former were much better. GA-SVM models can be useful for predicting binding modes and binding activities of the interactions of new antibiotic compounds with DNA.