DECRES

DECRES identifies enhancers and promoters in the human genome using supervised deep learning to map cis-regulatory elements for gene regulation studies.

Key Features:

• Deep Learning Framework: Built upon the Deep Learning Tutorials developed by LISA lab, DECRES employs neural network models to interpret complex genomic datasets.
• Data Integration: The method integrates annotations and functional data from ENCODE and FANTOM for analysis of non-coding regions.
• Genome-Wide Prediction: DECRES produces genome-wide predictions, reporting approximately 300,000 candidate enhancers (6.8% of the genome) and 26,000 candidate promoters (0.6% of the genome).
• Experimental Feature Identification: The approach analyzes well-characterized cell lines to identify experimental features that enhance predictive performance.
• Validation with Bidirectional Transcription Data: A subset of predicted enhancers is supported by bidirectional transcription evidence.

Scientific Applications:

• Gene Regulation Studies: Identification of active cis-regulatory regions to elucidate mechanisms of transcriptional control.
• Phenotypic Impact Assessment: Annotation of non-coding regions to evaluate how genetic variation may influence phenotypes.
• Functional Genomics: Provision of genome annotations for enhancers and promoters to support functional interpretation of regulatory elements.
• Clinical Applications: Predicted regulatory annotations to aid interpretation of non-coding variants in clinical genomics.

Methodology:

DECRES applies supervised deep learning models (based on the Deep Learning Tutorials by LISA lab) trained on large, well-characterized genomic datasets to learn patterns that distinguish enhancer and promoter regions.