Abstract 3578: Methyl2Acetyl: predicting epigenetic and transcriptomic activity from DNA methylation

Abstract

Abstract DNA methylation and histone modifications are interconnected and interdependent epigenetic mechanisms that regulate chromatin accessibility in cells. They establish cell’s developmental identity and modulate individual response to endogenous developmental stimuli and environmental changes. Chromatin immunoprecipitation followed by sequencing (ChIP-seq) is a popular assay to study genomewide histone modification profiles. However, ChIP-seq assays require relatively large amounts of starting materials (ranging from ~105 cells using a low cell number protocol to 107 cells in a standard protocol), which remains as a major obstacle for precious biological samples including primary tumor specimens. On the other hand, tagmentation-based whole genome bisulfite sequencing (TWGBS) protocol enables unbiased assessment of the genomewide methylation pattern using ~3,000 cells. Therefore, a quantitative relationship between DNA methylation and epigenetic/transcriptomic activity is in great demand. In the current study, we developed Methyl2Acetyl, a machine learning framework to infer promoter epigenetic activities (e.g. H3K27Ac enrichment) using TWGBS data as input. The proposed model utilizes random forest learners, which allow automatic extraction of complex interactions among DNA methylation features surrounding transcription start sites. We applied Methyl2Acetyl to a set of pediatric solid tumor samples with high quality H3K27Ac ChIPseq data to evaluate its performance. The model accurately predicted promoter H3K27Ac enrichment in independent test samples (Pearson correlation against measured enrichment: 0.81 ± 0.02). To further test the robustness of our model, we trained a Methyl2Acetyl model on solid tumors and applied it to a set of 10 pediatric leukemia samples. Despite dramatic global methylome differences between solid tumors and leukemia, Methyl2Acetyl predicted promoter H3K27Ac enrichment showed strong concordance with target gene expression level in the same sample (Pearson correlation: 0.75 ± 0.02, classification AUC: 0.89 ± 0.01). Collectively, our data suggested that DNA methylation is predictive of promoter active histone modification enrichment and gene expression activity. Epigenetically active promoters can be imputed from TWGBS data in samples with limiting starting materials. Citation Format: Xiang Chen. Methyl2Acetyl: predicting epigenetic and transcriptomic activity from DNA methylation [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3578. doi:10.1158/1538-7445.AM2017-3578