DeepRegFinder

DeepRegFinder identifies and classifies DNA regulatory elements such as enhancers and promoters at the genomic scale using convolutional and recurrent neural networks to enable genome-wide prediction from ChIP-seq data.

Key Features:

• Automated Workflow: Automates data processing, model training, and genome-wide prediction from ChIP-seq datasets.
• Neural Network Architecture: Employs convolutional and recurrent neural networks to capture complex sequence patterns indicative of enhancers and promoters.
• Classification of Regulatory States: Classifies identified enhancers and promoters into active and poised states.
• Genome-wide Prediction: Performs genome-wide identification and annotation of regulatory elements across multiple cell types.
• Performance Metrics: Quantifies performance using mean average precision and reports improvements relative to existing algorithms across cell types.
• Customizability: Allows customization for specific datasets and experimental conditions.

Scientific Applications:

• Gene Regulatory Network Analysis: Facilitates mapping of enhancers and promoters to study gene regulatory networks.
• Cellular Differentiation Studies: Supports investigation of regulatory element dynamics during cellular differentiation by distinguishing active and poised states.
• Disease Mechanism Research: Aids identification of regulatory element alterations relevant to disease mechanisms.
• Comparative and Evolutionary Genomics: Enables comparative analyses of regulatory elements across cell types and evolutionary contexts.

Methodology:

Automated data processing and training of convolutional and recurrent neural network models on known enhancer and promoter sites using ChIP-seq datasets, followed by application of trained models for genome-wide prediction and evaluation using mean average precision.