Feature Selection and Combination Criteria for Improving Accuracy in Protein Structure Prediction

Abstract

The classification of protein structures is essential for their function determination in bioinformatics. At present, a reasonably high rate of prediction accuracy has been achieved in classifying proteins into four classes in the SCOP database according to their primary amino acid sequences. However, for further classification into fine-grained folding categories, especially when the number of possible folding patterns as those defined in the SCOP database is large, it is still quite a challenge. In our previous work, we have proposed a two-level classification strategy called hierarchical learning architecture (HLA) using neural networks and two indirect coding features to differentiate proteins according to their classes and folding patterns, which achieved an accuracy rate of 65.5%. In this paper, we use a combinatorial fusion technique to facilitate feature selection and combination for improving predictive accuracy in protein structure classification. When applying various criteria in combinatorial fusion to the protein fold prediction approach using neural networks with HLA and the radial basis function network (RBFN), the resulting classification has an overall prediction accuracy rate of 87% for four classes and 69.6% for 27 folding categories. These rates are significantly higher than the accuracy rate of 56.5% previously obtained by Ding and Dubchak. Our results demonstrate that data fusion is a viable method for feature selection and combination in the prediction and classification of protein structure.

Fast SCOP Classification of Structural Class and Fold Using Secondary Structure Mining in Distance Matrix

“…These methods can be categorized into two groups in the case of assessment approach. The first group [5,7,19,20,27]applies the training set to build classifier model and use independently the testing set to evaluate its performance. The second group [16,21] combines these two sets and assesses the performance of classification by 10-cross validation.…”

“…The indirect representation can be organized into two types: based on statistical analysis of amino acid residues [5][6][7][8], and based on amino acid indices [9,10]. Another approach executes directly analysis on protein spatial structure to obtain representation and extract feature of structure, and can be grouped into three types: based on spatial atom distribution [11,12], topological structure [13,14], and geometrical shape [15][16][17].…”

“…The indirect representation can be always obtained with lower computation cost and higher dimensional feature, but in contrast the direct representation can be acquired with higher computation cost and lower dimensional feature. Besides, in order to achieve better classification with the indirect representation of protein structure, there is always a need to exploit advance techniques of pattern recognition, for example, feature combination [5], fusion [18,19], selection [7] and hierarchical classifier architecture [20,21].…”

Protein classification using texture descriptors extracted from the protein backbone image

“…(Chinnasamy et al, 2005;Shi et al, 2006) are based on the same features CSHPVZ but different classifier systems. (Huang et al, 2003) combines CSHPVZ with bigram-coded feature (B) and spaced bigram-coded feature(SB), (Lin et al, 2007) does the same work as (Huang et al, 2003) but improves the classifier system by the technique of data fusion.…”

“…Method Accuracy (%) (Ding & Dubchak, 2001) 56.50 (Chinnasamy et al, 2005) 58.18 (Shi et al, 2006) 61.04 (Huang et al, 2003) 65.50 (Lin et al, 2007) 69.60 (Shi & Zhang, 2009) 72.99 FUS3 81.04 Table 4. Other comparisons in the FOLD dataset.…”