piSAAC: Extended notion of SAAC feature selection novel method for discrimination of Enzymes model using different machine learning algorithm


Abstract in English

Enzymes and proteins are live driven biochemicals, which has a dramatic impact over the environment, in which it is active. So, therefore, it is highly looked-for to build such a robust and highly accurate automatic and computational model to accurately predict enzymes nature. In this study, a novel split amino acid composition model named piSAAC is proposed. In this model, protein sequence is discretized in equal and balanced terminus to fully evaluate the intrinsic correlation properties of the sequence. Several state-of-the-art algorithms have been employed to evaluate the proposed model. A 10-folds cross-validation evaluation is used for finding out the authenticity and robust-ness of the model using different statistical measures e.g. Accuracy, sensitivity, specificity, F-measure and area un-der ROC curve. The experimental results show that, probabilistic neural network algorithm with piSAAC feature extraction yields an accuracy of 98.01%, sensitivity of 97.12%, specificity of 95.87%, f-measure of 0.9812and AUC 0.95812, over dataset S1, accuracy of 97.85%, sensitivity of 97.54%, specificity of 96.24%, f-measure of 0.9774 and AUC 0.9803 over dataset S2. Evident from these excellent empirical results, the proposed model would be a very useful tool for academic research and drug designing related application areas.

Download