Study On Adjuvant Diagnosis And Staging Of Lung Adenocarcinoma Based On Computational Methods And DNA Methylation

Studying disease using bioinformatics has become the greatest interested topic for researchers with the development of bioinformatics.DNA methylation is one of the most important chemical modifications in epigenetics which playing a key role in the gene regulation and close related to various physiological processes.Transcriptional inactivation of tumor suppressor genes and proto-oncogene expression caused by abnormal methylation of Cp G islands are suspected to be the main factors leading to cancer.Compared with conventional invasive methods to diagnose cancer,understanding the relationship between DNA methylation and cancer makes it possible to diagnose,stage and prognosis of cancer invasively,providing Compared with conventional invasive methods to diagnose the cancer process,understanding the relationship between DNA methylation and cancer makes it possible to diagnose,stage and prognosis of cancer without invasive methods,providing a feasible way to judge the treatment results immediately.Lung adenocarcinoma(LUAD)was taken to conduct this research.We downloaded 450 K methylation microarray data and clinical data from TCGA database,UCSC Xena website and GEO database;using R software,Ch AMP package to analyze our data preliminary,providing a statistical support for the diagnosis of LUAD.In our LUAD prediction experiment(chapter 4),we analyzed the difference methylation position(DMP)between tumor group and control(healthy)group,and constructed 9 dataset of different number of probes.DNN and RF were taken to predict LUAD respectively,5-Folding cross validation has achieved good classification results.(AUC > 0.999).The experimental results verified the rationality of selecting ? value as biomarker in diagnosis of LUAD,and provided a good idea for the early screening and diagnosis of lung adenocarcinoma and real-time efficacy evaluation.Additionally,we also get a table about 70 most differential genes between LUAD and healthy people,among them PTPRN2,PRDM16,TNXB,ADARB2 and RASA3 were most different.Finally,we constructed an unbalanced data set to predict cancer staging(early:intermediate: late = 12:9:1),using SMOTE to deal with the imbalance of cancer staging data sets,and proposed FULL-Stage,which got a satisfier score(ACC=76.493,PRE=71.038,REC=68.762,MCC=55.152)compared with SVM and RF,this shows that our proved algorithm has a better generalization ability.This study demonstrated that abnormal DNA methylation do closely related to lung adenocarcinoma(LUAD),and the satisfactory results suggested a new method and idea in LUAD diagnosis,which may be practical to assist medical practitioners on their clinical work.Through this study,we expect this method will be helpful for the study of diagnosis and staging in other cancers and other methylation related diseases.