Function Research And Biomarker Identification Of Nervous System In Cancer

Research shows that nerves are important pathological elements of the tumor microenvironment.The relationship between nerves and cancer is a hot spot and trend in current cancer research.At present,there are few reports on mining the relationship between nerves and cancer through machine learning methods.This paper collects4039 neural genes based on the literature,and then uses machine learning and statistical methods to explore the function of the nervous system in cancer and identify neural biomarkers of cancer metastasis,and then establish an online cancer neural function analysis platform for researchers.To provide some theoretical support for the research on the relationship between nervous system and cancer.Specifically,the main contents of this paper are as follows:（1）Neural biomarker identification of cancer metastasis.This paper identified neural biomarkers of cancer metastasis in nine types of cancer（Adrenocortical carcinoma（ACC）,Breast invasive carcinoma（BRCA）,Colon adenocarcinoma（COAD）,Kidney renal clear cell carcinoma（KIRC）,Kidney renal papillary cell carcinoma（KIRP）,Lung adenocarcinoma（LUAD）,Rectum adenocarcinoma（READ）,Stomach adenocarcinoma（STAD）,Thyroid carcinoma（THCA））.Stages I and II of cancer are regarded as the non-metastatic group,stage IV is regarded as the metastatic group,and stage III is in a transitional state.Some samples have lymph node spread,and some samples do not,so this paper does not consider it in the experiment,which also leads to the problem of category imbalance.This paper uses SMOTE（Synthetic Minority Oversampling Technique）to deal with this problem.This paper compares nine feature selection algorithms,such as T-test,and seven classification algorithms,such as random forest.First,the features were selected by T-test on KIRC and COAD,and seven classification algorithms were evaluated,with the increase of the number of features,the accuracy of random forest has always been better.Therefore,this paper uses the random forest classification model in the following experiments.The experimental results show that the performance accuracy of the nine cancer metastasis diagnosis models using the neural gene dataset has reached more than 90%,which indicates that neural genes play a role in nine types of cancer metastasis.The paper also found that the optimal model using neural gene datasets in BRCA,KIRP,LUAD cancers achieved a classification accuracy of almost 1.00,which was higher than the accuracy of random feature selection,suggesting that neural genes play a significant role in BRCA,KIRP,LUAD cancer metastasis.（2）Studies have shown that gender affects cancer biomarkers,especially genes related to tumor aggressiveness,and that patients of different ages also have different rates of metabolism and cell development.Therefore,this paper investigates whether neural genes can better identify cancer metastasis by combining gender and age information,respectively.In the above nine data sets,KIRC combines gender and age information,and the sample is more comprehensive.Therefore,this subsection selects KIRC for research.The KIRC data set is divided into male and female,and compared with all genders;the samples in the KIRC data set that are older than or equal to 60 are divided into the old group,and the samples less than 60 years old are divided into the young group,and compared with the whole age.Spearman’s correlation coefficient results did not show a significant correlation between cancer stage and gender or age.The results of this paper show that the evaluation results of the nine feature selection algorithms all support that the accuracy of the random forest classification model on the female dataset is much better than the accuracy of the classification model on the full gender dataset,the accuracy of the classification model for the male dataset is slightly worse than or similar to the accuracy of the classification model for the full gender dataset.The accuracy of the classification model for the young dataset is much better than that of the full-age dataset,and the accuracy of the older dataset is slightly worse than or similar to that of the full-age dataset.（3）This paper delves into the function of neural genes in 26 cancers.The neural genes in cancer were analyzed from four aspects: expression analysis of neural genes in cancer,survival analysis,co-expression network analysis,and correlation analysis of neural genes and non-neural genes in tumor microenvironment.The results of the study show that the neurological functions of cancers with high survival rate and those with low survival rate are different;differentially expressed neural genes have a great influence on the prognosis of patients,and most of the effects are related to poor prognosis;The more malignant cancer,the more active the communication between neurons;the nerves can help cancer cells overcome the stress in the tumor microenvironment.（4）This paper provides an online analysis platform（Neural Genes in Cancers,NGC）for cancer researchers.Researchers can study the contribution of nerves to cancer occurrence,development,and metastasis at the pan-cancer level.At the same time,it also includes some online analysis tools,such as the online pathway enrichment applet.Users can obtain information on the expression,survival,correlation,co-expression of neural genes in cancer,and expression profiles of neural genes in the development of different organs.Taken together,our work not only reveals the remarkable functions of the nervous system in cancer initiation,progression,and metastasis,but also identifies potential neural biomarkers of cancer metastasis and provides biologists with an online analysis tool,realize the online analysis of neural function in pan-cancer.