MATEPRED-A-SVM-Based Prediction Method for Multidrug And Toxin Extrusion (MATE) Proteins. (October 2015)

• Record Type: Journal Article
• Authors: Tamanna, ; Ramana, Jayashree
• Is Part Of: Computational biology and chemistry. Volume 58(2015)
• Page Start: 199
• Page End: 204
• Publication Date: 2015-10
• Subjects: Antibiotics -- Drug resistance -- MATE -- PSSM -- SVM -- Diarrheal pathogens; Chemistry -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; Biochemistry -- Data processing; Biology -- Data processing; Molecular biology -- Data processing; Periodicals; Electronic journals; 542.85
• DOI: 10.1016/j.compbiolchem.2015.07.011
• Languages: English
• ISSNs: 1476-9271

Abstract

Graphical abstract: Highlights: The prediction of MATE proteins is made using SVM based method. The model was derived using PSSM classifier and achieved an overall accuracy of 92.06%. MATEPred efficiently distinguishes MATE proteins from non-MATE proteins. This tool is helpful to general users and also for the researchers working in this field. Abstract: The growth and spread of drug resistance in bacteria have been well established in both mankind and beasts and thus is a serious public health concern. Due to the increasing problem of drug resistance, control of infectious diseases like diarrhea, pneumonia etc. is becoming more difficult. Hence, it is crucial to understand the underlying mechanism of drug resistance mechanism and devising novel solution to address this problem. Multidrug And Toxin Extrusion (MATE) proteins, first characterized as bacterial drug transporters, are present in almost all species. It plays a very important function in the secretion of cationic drugs across the cell membrane. In this work, we propose SVM based method for prediction of MATE proteins. The data set employed for training consists of 189 non-redundant protein sequences, that are further classified as positive (63 sequences) set comprising of sequences from MATE family, and negative (126 sequences) set having protein sequences from other transporters families proteins and random protein sequences taken from NCBI while in the test set, there are 120 protein sequences in all (8 in positive and 112 in negative set). The model was derived using Position Specific Scoring Matrix (PSSM) composition and achieved an overall accuracy 92.06%. The five-fold cross validation was used to optimize SVM parameter and select the best model. The prediction algorithm presented here is implemented as a freely available web server MATEPred, which will assist in rapid identification of MATE proteins.