Genome-wide prediction of cis-regulatory regions using supervised deep learning methods

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• The human genome contains 98% non-coding DNA, previously considered junk, which harbors crucial cis-regulatory elements controlling gene expression.
• Identifying active cis-regulatory regions is vital for understanding gene regulation and the effects of genetic variations on phenotypes.
• Advancements in high-throughput sequencing and machine learning enable genome-wide prediction of these regulatory regions.

Purpose of the Study:

• To develop a deep learning model (DECRES) for accurate identification of enhancer and promoter regions in the human genome.
• To leverage large-scale datasets from ENCODE and FANTOM projects for cis-regulatory element prediction.
• To advance the understanding of genomic locations and functions of cis-regulatory regions.

Main Methods:

• Utilized supervised deep learning approaches for cis-regulatory region identification.
• Employed data from Encyclopedia of DNA Elements (ENCODE) and Functional Annotation of the Mammalian Genome (FANTOM) projects.
• Developed predictive models using well-characterized cell lines to identify key experimental features.

Main Results:

• DECRES identified approximately 300,000 candidate enhancers (6.8% of the genome) and 26,000 candidate promoters (0.6% of the genome).
• 40,000 of the identified enhancers were supported by bidirectional transcription data.
• The deep learning model successfully identified patterns in complex genomic data for predicting cis-regulatory regions.

Conclusions:

• Predicted annotations of cis-regulatory regions offer significant utility for genome interpretation in functional genomics and clinical applications.
• The DECRES model highlights the potential of combining deep learning with high-throughput sequencing data for genome annotation.
• This work inspires the development of advanced neural network models for improved genome annotation accuracy.