A Novel Local And Global Similarity Based Feature Extraction Approach For Protein Classification

2872 Words Sep 30th, 2014 12 Pages
A Novel Local and Global Similarity based Feature
Extraction Approach for Protein Classification
Abstract—In this article, a novel approach is proposed based on local and global similarity for extracting features from protein sequences. The proposed approach extract only 6 features corresponding to each protein sequence. These features are computed by globally considering the probabilities of occurrences of the amino acids in different position of the sequences which locally belongs to the six exchange groups [1]. Then, these features are used as an inputs for Neural Network learning algorithm named as Boolean-Like Training Algorithm (BLTA) [2]. The
BLTA classifier is used to classify the protein sequences obtained from the Protein Information Resource (PIR) maintained by the
National Biomedical Research Foundation (NBREF-PIR) [3]. To investigate the efficacy of proposed feature extraction approach, the experimentation is performed on two superfamilies, namely
Ras and Globin. Across tenfold cross validation, the highest
Classification Accuracy and Computational Time achieved by proposed approach is 94.323.52 and 6.54(s) respectively in comparison to the Classification Accuracies achieved by other approaches
[4], [5] and [6] are 85.420.55, 67.518.38, 51.410.27 with Computational Time 7.11(s), 10.13(s), 63.98(s) respectively.
The experimental results demonstrate that the proposed approach extract the minimum relevant features for each protein sequence.
Therefore, it…
Open Document