Machine Learning Approach For Computational Analysis Of Human UTRS and Cancer Causing Genes To Find Correlations Amongst Various Genomic Features And Development Of Its Predictive Model

Abstract

ABSTRACT newlineCancer is a group of diseases involving abnormal cell growth with the newlinepotential to invade or spread to other parts of the body. Although not all tumors are newlinecancerous; benign tumors do not spread to other parts of the body. Possible signs newlineand symptoms include a lump, abnormal bleeding, prolonged cough, unexplained newlineweight loss and a change in bowel movements. While these symptoms may indicate newlinecancer, they may have other causes also. More than hundred types of cancers newlineaffect humans. These cancers are actually related with over expression of the newlinevarious metabolic genes. If we go in to the biological system in depth than we will newlineknow that UTR (Untranslated part of the gene) is the main regulator of these over newlineexpressed genes. These are the true creators of the cancer causing genes. The newlinesequence vectorization followed by the machine learning methods provides a newlinedetailed insight into the UTR composition and prevalence of sequence features for newlinethe classification of UTRs. In the current study we have built a customized machine newlinelearning model using various monomeric, dimeric, trimeric, entropy, AT and GC newlinefeatures which helps to classify the UTR sequences into cancerous and noncancerous 5 to 3 classes. In current study various classification algorithms were newlinecompared and Random Forest Algorithm was found the best amongst them. The newlinecurrent model has provided an accuracy of 91% to classify the 3 and 5 classes newlinerespectively. All the mutual correlations of the extracted features were taken in newlineconsideration for the proper analysis of 3 and 5 UTR and this statistical study also newlineindicating the correlation of UTR features with the spread/intensity of cancer. This newlinedata analysis helped a lot to classify diseased/Non-diseased genes . The newlinecomprehensive study also helped in determining the expected evolutionary change newlinethat may occur in gene during the course of evolution by finding the evolutionary load newlineon the genes. newline