Classification of Mutated Cancer Genome Using Machine Learning Approaches

Abstract: Cancer may be a genetic abnormality derived from genetic changes that end in a loss of control over necessary cellular functions. Identification of mutated cancer gene plays a crucial role in individualizing the treatment of a cancer patient in keeping with his specific tumorigenic profile. Wavelets analysis techniques are capable of extracting each spectral and local information and perform multiscale analysis on DNA/protein sequences. The amino acid index features represent the physicochemical properties of the protein sequences. The wavelet features combined with AAIndex features offer feature vector for classification of mutated driver gene. Machine learning based approaches are utilized in cancer genome analysis to mine patterns from the prevailing data and built mathematical models to learn patterns and make predictions in unanalyzed data. The proposed system deals with scrutiny the performance of varied combinations of Support Vector Machine and Back Propagation Neural Network to identify the mutated driver gene.

Keywords: driver gene; wavelet analysis; AAIndex features; Support Vector Machine ; Back Propagation Neural Network.