Identification of new prognostic DNA methylation features in uveal melanoma by 11+ based on methylation group and transcriptome analysis~

Introduction

Uveal melanoma （UVM） It is the most common primary intraocular human malignant tumor , The death rate is very high . abnormal DNA Methylation has rapidly become a diagnostic and prognostic feature of many cancers .

Background introduction

Gene methylation plays an important role in tumorigenesis and development , This article is brought to you by Xiaobian today , Genome - wide integration analysis of methylosome and transcriptome was carried out , A strategy based on machine learning is proposed to recognize 10 individual Methylation driven prognostic genes (MDPG) Composed of DNA Methylation drive flag (10MeSig) . Published in 《Briefings in Bioinformatics》 On , The influence factor is 11.622, The title of the article is ：Machine learning-based integrative analysis of methylome and transcriptome identifies novel prognostic DNA methylation signature in uveal melanoma.

Data is introduced

The data used in this study ： from UCSC Xena Downloaded TCGA Medium 80 individual UVM Clinical information of cases 、 Methylation and transcriptome data ; from GSE44295 Other independence gained in 57 individual UVM Case data .

Result analysis

01

MDPGs Identification of

The workflow of this study is shown in the figure 1 Shown . The first 1 Step ： The methylation group and transcriptome in the top solid frame were identified by comprehensive analysis MDPG. The first 2 Step ： Establish methylation driven prognostic features using machine learning methods . The first 3 Step ： Validation of methylation driven prognostic features in an independent cohort .

chart 1

In this study, differential methylation analysis was performed in early and late stage patients , A total of 3982 Stage related CpG site . among , Find out 2187 Candidate prognosis CpG Significantly correlated with overall survival （ chart 2A）. Manhattan map shows , These related loci are unevenly distributed in human autosomes （ chart 2B）. Next, this study is about 2187 Candidate prognosis CpG Unsupervised hierarchical cluster analysis of loci produced two different patient clusters , They are significantly correlated with clinical staging （ chart 2C）. And then 2187 Candidate prognosis CpG When mapping loci to genomic compartments , This study found that most of them are rich in promoter and genomic regions , involve 304 Highly methylated genes and 619 Hypomethylated genes （ chart 2D and E）, Most belong to the coastal and high seas regions of the genome （ chart 2F）.

To identify genes related to progress , This study also analyzed the differential expression between early and late stage patients and between metastatic and non metastatic patients . Results a total of 219 Stage related genes and 1013 Metastasis related genes （ chart 2G and H）. Last , This study will 40 Hypermethylated and hypomethylated genes significantly associated with stage or metastasis were defined as MDPG（ chart 2I）.

chart 2

02

MDPGs Functional representation of

To explore 40 individual MDPG The function of , This study on MDPG the GO and KEGG Pathway function enrichment analysis .GO The analysis results are shown in the figure 3A Shown ,40 individual MDPGs Enriched in 7 individual GO BP in ,KEGG The analysis results are shown in the figure 3B Shown ,40 individual MDPG Participate in protein digestion and absorption and axon guidance .

chart 3

For further study MDPGs And UVM The relationship between clinical features , This study is based on 40 The expression pattern of genes is right TCGA In dataset 80 name UVM The patients were analyzed by consistent cluster analysis , according to CDF Three different patient clusters were obtained from the regional changes （ chart 3C and D), There were significant differences in overall survival among the three patient clusters （ chart 3E）.

03

MeSig Development and verification of

This study uses an enhanced machine learning algorithm from 40 individual MDPG Select a signature in and identify a MeSig, Include 10 individual MDPG（10MeSig）. then , take 10MeSig adopt 10MeSig The linear combination of the expression is transformed into a risk scoring model , And through multiple Cox The relative coefficients in the regression are weighted .10MeSig The optimal risk threshold will be 80 Patients were divided into high-risk groups with different overall survival (n = 30) And the low-risk group (n = 50). Pictured 4A Shown , The overall survival rate in the high-risk group was significantly lower than that in the low-risk group . Time dependence ROC Analysis and display ,10MeSig forecast 3 Years and 5 Years of overall survival AUC Respectively 0.95 and 1.0（ chart 4B）.

chart 4

For further study DNA Methylated pair UVM Effects of gene regulation , Firstly, we study 10MeSig Press 10MeSig Expression and... In stratified high-risk and low-risk groups DNA Methylation patterns . Pictured 5A Shown , be based on DNA The intersection between methylation and gene expression , The observed 10MeSig The four expressions and DNA Methylation patterns . Next, this study passed Spearman Correlation analysis further tested DNA The effect of methylation on gene expression , Find out 7 individual CpG- Gene pairs showed significant negative correlation ,4 individual CpG- Gene pairs showed significant positive correlation （ chart 5B）.

chart 5

04

10MeSig Independent verification of

For further verification MDPGs And UVM The relationship between clinical features , This study was conducted in an independent GEO The data is plotted 35 individual MDPGs gene . then , Yes GEO Queue 57 name UVM The patients were analyzed by consistent cluster analysis , from 35 The expression patterns of these genes reveal two different patient clusters （ chart 6A and 6B）. Pictured 6C Shown , There was a significant difference in overall survival between the two patient clusters ; Pictured 6D Shown , The overall survival time of patients in the high-risk group was significantly shorter than that of patients in the low-risk group . Besides ,9 individual MDPG The expression pattern of is consistent with that observed in the discovery cohort （ chart 6E）.

chart 6

05

10MesSig It has nothing to do with clinical and pathological features

For further examination 10MeSig Whether it is independent of other clinical and pathological factors , This study conducted a multifactor study Cox regression analysis . Pictured 7A Shown , stay TCGA-UVM In line ,10MeSig , IV Period and age were correlated with OS significant correlation . In the independent GEO Of the queue , Only 10MeSig And OS Maintain significant correlation （ chart 7B）. These results suggest that ,10MeSig yes UVM Independent predictors of patient survival .

chart 7

Editor's summary

In this study , The authors performed genome-wide integration analysis of methylosome and transcriptome , To determine the 40 Methylation driver genes , And in order to further study this 40 Clinical application of methylation driving genes , The author makes feature selection based on machine learning , It is determined that 10 individual MDPG Composed of MeSig (10MeSig), It will TCGA Queued UVM Patients were stratified into two risk groups , Overall survival was significantly different . The research method of this paper is worthy of our study .