OligoPred: A web-server for predicting homo-oligomeric proteins by incorporating discrete wavelet transform into Chou's pseudo amino acid composition

In vivo, some proteins exist as monomers (single polypeptide chains) and others as oligomers. Not like monomers, oligomers are composed of two or more chains (subunits) that are associated with each other through non-covalent interactions and, occasionally, through disulfide bonds. These proteins are the structural components of various biological functions, including cooperative effects, allosteric mechanisms and ion-channel gating. However, with the dramatic increase in the number of protein sequences submitted to the public data bank, it is important for both basic research and drug discovery research to acquire the possible knowledge about homo-oligomeric attributes of their interested proteins in a timely manner. In this paper, a high-throughput method, combined support vector machines with discrete wavelet transform, has been developed to predict the protein homo-oligomers. The total accuracy obtained by the re-substitution test, jackknife test and independent dataset test are 99.94%, 96.17% and 96.18%, respectively, showing that the proposed method of extracting feature from the protein sequences is effective and feasible for predicting homo-oligomers. The online service is available at http://bioinfo.ncu.edu.cn/Services.aspx.

. Based on different basis functions, the wavelets have different families; every family has its unique quality fitting for certain signals. As the characteristics of the analyzing wavelet influence the performance of DWT, the better the analyzing wavelet matches the underlying structure in the signal, the better feature values can be extracted from the sequences. Therefore, selection of a suitable wavelet basis which possesses desirable properties such as compactly support, orthogonality, symmetry, smoothness and high order of vanishing moments is necessary for the signal processing. Nine wavelet functions were chosen for investigating the effect of the different wavelets on classification accuracy in the research. The R1568 database prediction results performed by different wavelet functions with Kyte–Doolittle hydrophobicity scales were listed in Fig. 1. As can be seen from Fig. 1, in jackknife test by using SVM, the training accuracy can reach 96.17% when using Bior2.4 wavelet to extract feature. However, when using other wavelet functions, the training accuracy only ranged from 84.82% to 92.66%. This is because wavelet of Bior2.4 is a class of symmetric and biorthogonal wavelets that have good ability to reduce dimension by Mallat decomposition and can remove high-frequency noises by selecting several or even one approximation sub-band. Therefore, Bior2.4 wavelet was selected as the appropriate wavelet function to input SVM to predict homo-oligomeric proteins in this study.Figure optionsDownload high-quality image (86 K)Download as PowerPoint slideHighlights
► Discrete wavelet transform (DWT) can effectively grasp the core features of the protein homo-oligomers.
► A new method, in which SVM combines with DWT, is developed to predict the homo-oligomers.
► The proposed approach can remarkably improve the predictive accuracies of the homo-oligomers.