Machine Learning Framework for Identification of Mutation in Gene Sequences to Detect Cancer Progress

Abstract

This study proposes a framework that will help in identifying the progression of cancer. Stomach carcinoma driver mutated genes code expression regulated pRpApApPpIpVp,p ApApPpIVp, FpDpV, PpRpIpM, RpPpRpIpMp in Central, Raw and Hahn moments to extractp Extensivep Featurep Vectors (EFVs) explained in feature extraction framework. Three variants of Recurrent Neural Network i.e., LpSpTpM, GpRpU, andp BiLSTM accumulated to comprehensively scrutinize the EFVs. Input sequence shape consists of 64, 1; reduced by Principal Component Analysis (PCA). In-depth analysis of defined models according to their performance and efficiency is portrayed as graph plots, confusion matrix, loss, and accuracy matrices. The GRU & BiLSTM shows best prediction metrics with accuracy of 99.46%p, psensitivity of 98.93%, pspecificity of 100%, pMCC as 0.989 andp pAUC of 1.00. The future work may lead to learning to improve framworks of various carcinomas that will lead the identification of progression cancer in extremely cheap prices with fastest results.