Identification of DNA-Binding Proteins via Fuzzy Multiple Kernel Model and Sequence Information

Abstract

DNA-binding proteins is the molecular basis for understanding the basic processes of life activities. Many diseases are associated with DNA binding proteins. The methods of detecting DNA-binding proteins are mainly realized by biochemical experiment, which is time consuming and extremely expensive. A lot of computational methods based on Machine Learning (ML) algorithm have been developed to detect DNA-binding proteins. In this study, we propose a novel DNA-binding proteins model via a Fuzzy Multiple Kernel Support Vector Machine. The multiple features of sequence and evolutionary are extracted and constructed as multiple kernels, respectively. Next, these corresponding kernels are integrated by Multiple Kernel Learning (MKL) algorithm. At last, Fuzzy Support Vector Machine (FSVM) is employed to build an effective DNA-binding protein predictor. Comparing with other outstanding methods, our proposed approach achieves good results. The accuracy of our model are 82.98% and 81.70% on PDB1075 (benchmark data set of DNA-binding proteins) and PDB186 (independent test set), respectively. Our approach is comparable to previous methods.