Classification of anti-oxidant proteins using novel physiochemical and conjoint-quad (PCQ) feature composition

The anti-oxidant proteins have a closer relation to disease control. Hence an accurate classification of antioxidant proteins by automated analysis is an essential process for the expansion of drugs for various diseases. Wet-lab experimental approaches are generally expensive and inefficient for the identification of anti-oxidant proteins. Novel methodologies like Physiochemical and Conjoint-Quad (PCQ) feature composition using the physio-chemical features combined with moment-based features is proposed in this work for the accurate classification of anti-oxidant proteins. In this proposed work, four techniques namely, proposed PCQ, k-spaced Amino Acid Pairs (CKSAAP), g-gap, and N-gram (N = 3) were applied to create different hybrid features from the anti-oxidant proteins efficiently. The Pearson Kernel-based Supervised Principal Component Analysis (PKSPCA) is proposed for the dimension reduction of the features and effective classification. To evaluate the proposed technique, ten-fold cross-validation and independent test datasets were utilized. On the testing data, the proposed method attained the best performance when compared with the previous techniques. This proposed method achieves 99% accuracy, sensitivity of 98% and specificity of 91% during the classification of the anti-oxidant proteins.

Automated summary

Accurate classification of antioxidant proteins is crucial for the development of drugs for various diseases. This study proposes a novel methodology using feature composition and dimension reduction techniques for the accurate classification of antioxidant proteins.