Abstract
Motivation Enhancers and promoters are important classes of DNA regulatory elements that control gene expression. Identifying them at the genomic scale is a critical and challenging task in bioinformatics. The most successful method so far is to train machine learning models on known enhancer and promoter sites and predict them at other genomic regions using ChIP-seq and related data.
Results We have developed a highly customizable program called DeepRegFinder which automates data processing, model training and genome-wide prediction of enhancers and promoters using convolutional and recurrent neural networks. Our program further classifies the enhancers and promoters into active and poised states to facilitate downstream analysis. Based on mean average precision scores of different classes across multiple cell types, our method significantly outperforms the existing algorithms.
Availability https://github.com/shenlab-sinai/DeepRegFinder
Competing Interest Statement
The authors have declared no competing interest.