Cancer Classification Based On Gene Expression And DNA Methylation Data

Cancer is a deadly disease all over the world and its morbidity is increasing at an alarming rate in recent years,especially in China.With the rapid development of computer science and machine learning technologies,computer-aided cancer detection has achieved increasingly progress.In recent years,genome sequencing technology has made significant breakthroughs,more and more omics datas of cancer tissues have been generated.Nowadays,using machine learning methods to mine the omics datas for understanding their relationship with the formation and progression of cancer has become a research hotspot.Starting with RNASeq data and DNA methylation data,this paper aims to explore the application of machine learning and deep learning methods in the field of cancer omics datas mining.For RNA-Seq data,we build a deep convolutional neural network for cancer diagnosis and cancer type prediction.Experimental results on three types of cancer datasets of Breast Cancer,Lung Adenocarcinoma and Stomach Cancer show the proposed deep convolutional neural network based method can extract more abstract features from input data through multilayer nonlinear transformation and achieve better prediction results,precision and recall are also improved.For DNA methylation data,although deep convolutional neural networks have superiority in automatic feature extraction,due to the high dimension of samples feature space and small scale clinical datasets,it is easily over-fitting of using a large scale convolutional neural networks to deal with the original input data.To address this problem,in this paper,we build a convolutional neural network based ensemble method for cancer prediction.Experimental results on three types of cancer datasets of Lung Adenocarcinoma,Liver Hepatocellular Carcinoma and Kidney Clear Cell Carcinoma show the effectiveness of the machine learning method in DNA methylation data mining.The proposed convolutional neural network based ensemble method combines the advantages of traditional machine learning methods and deep convolutional neural networks,the prediction results and several performance indexes are also improved.In brief,experimental results show that the machine learning and deep learning methods have the ability of RNA-Seq and DNA methylation data mining and thus providing a new insight for cancer diagnosis.