A Statistical Approach To Tumor Methylation Profile Purification

The high heterogeneity of tumor tissue is an important issue in cancer genomics.Due to the high heterogeneity of tumor tissue,methylation profiles of tumor samples obtained in clinical experiments are always mixture signals from different components,including cancer?normal?stromal cells?and Immune cells,etc.Among them,the admixture of normal cells is deemed as a major confusing factor for many downstream analyses.Neglection or improper consideration of tumor purity could bring biased or error results in DNA methylation analysis.Therefore,it is very important to establish an appropriate statistical model to correct the purity of tumor.Here we developed a linear statistical model,InfiniumPurifyMT,to purify cancer methylomes from tumor samples,adjacent normal samples and tumor purities.It can be treated as a modified version of our previous method InfiniumPurify by subdividing the error into two parts that one part comes from cancer samples,and the other comes from normal samples.We systematically investigated the accuracy of our method compared with InfiniumPurify on simulation and real data.Our modified method was shown to have better performance under different scenarios,including different number of matched normal samples,perturbation of purities and noises of matched normal methylomes and InfiniumPurifyMT shows better performance in different situations.Finally,in the real data,we took four cancer types BRCA,BLCA,LUAD and CHOL as examples to analyze the functional enrichment of differentially methylated genes obtained only by InfiniumPurifyMT.Result from our method is more accurate and biologically meaningful than InfiniumPurify.