Bio-M: Data mining on HCV genotype 1 core sequences

Hepatitis C Virus (HCV) has become a major risk factor for the development of Hepatocellular Carcinoma (HCC). A framework has been developed to identify genomic markers associated with HCC of HCV sequences, which comprises of clustering, feature selection and classification. A new method for feature extraction for genomic sequences rooted in Hash tables has been proposed. It requires less memory compared to Generalized Suffix Tree based methods. Biomarkers are selected as features and Random Forest (RF) Classifier is learned by means of these biomarkers. RF is used to classify HCV sequences with and without HCC. Using the HCV sequence data available from the European HCV Database (euHCVdb) and Los Alamos National Laboratory, we show that performance of RF is comparable with SVM classifier.