What is it about?
We describe Methylphet, a machine learning method to predict transcription factor (TF) binding in vivo based on the methylation profile measured by whole genome bisulfite sequencing (WGBS) studies. Using Methylphet is like perform a virtual ChIP-seq for any TF as long as you have the methylation profiles from WGBS on that cell type or clinical sample.
Why is it important?
It is difficult to do ChIP-seq on clinical samples because there is just not enough cells. Even if you can, ChIP-seq is an expensive experiment that heavily depends on the quality of the antibody. Our work showed that instead of doing ChIP-seq, you can instead conduct WGBS on that sample (which works well on very limited material like clinical samples), and computational predict the in vivo binding sites of a TF genome-wide. Our work is partially inspired by CENTIPEDE, which use DNase profile to predict TF binding in vivo, but because we use supervised learning (CENTIPEDE use mixture model, an unsupervised method), our results are much more stable and superior than CENTIPEDE in most of the datasets we tested. A side finding is that not only CG methylation, but also CH methylation is correlated with TF binding,
Read the Original
This page is a summary of: Base-resolution methylation patterns accurately predict transcription factor bindings in vivo, Nucleic Acids Research, February 2015, Oxford University Press (OUP), DOI: 10.1093/nar/gkv151.
You can read the full text:
The following have contributed to this page