Differential methylation site identification method

Abstract

The invention discloses a differential methylation site identification method. According to a thought of classification, identification for differential methylation sites is converted to searching of sites having important contributions to classification by a feature selection approach, and the sites having important contributions to classification are differential methylation sites. Specifically the differential methylation site identification method includes: firstly preprocessing 450K methylation chip data acquired from a public data standardizing the 450K methylation chip data to eliminate within-block errors, eliminating batch effects to eliminate inter-group errors and eliminating the sites small in variances; secondarily, constructing a random forest model to obtain contribution values of every site to classification; finally, determining the site as the differential methylation site if the contribution value of the site is larger than 0. The differential methylation site identification method has the advantages that the obtained differential methylation sites have better class judging performances and can provide more accurate results for cancer diagnosis.

Description

technical field [0001] The invention belongs to the technical field of DNA methylation recognition, and more specifically relates to a differential methylation site recognition method. Background technique [0002] As the most typical epigenetic phenomenon in the human genome, DNA methylation plays an important role in many key physiological activities. Its methylation status is closely related to the occurrence of various diseases, especially cancer. Specifically, not all methylation sites are associated with cancer, only some specific methylation sites are associated with cancer. These specific methylation sites are called differential methylation sites in this paper. [0003] At present, statistical methods are usually used in differential methylation site identification algorithms, such as fastDMA using variance analysis, and ChAMP using linear regression combined with t hypothesis testing methods. However, using traditional statistical methods can only roughly find s...

AI Technical Summary

Problems solved by technology

However, using traditional statistical methods can only roughly find sites with statistically significant differences, so many differentially methylated sites are found, and not all differentially methylated sites have the function of cancer diagnosis

Images