Computational methods for identifying enhancer-promoter interactions.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Enhancer-promoter interactions (EPIs) are key components of the human genome's cis-regulatory mechanism.
• Identifying EPIs is essential for understanding gene expression regulation.
• Current methods for detecting EPIs require a systematic review to aid researchers in application and optimization.

Purpose of the Study:

• To provide a comprehensive review of methods for identifying enhancer-promoter interactions (EPIs).
• To describe a framework for predicting EPIs and summarize available datasets and prediction tools.
• To review the application of EPI identification methods in disease contexts, particularly cancer.

Main Methods:

• Systematic review of sequencing technologies and computational models for EPI identification since 2010.
• Categorization of prediction methods based on data features (genetic, genomic, epigenomic).
• Evaluation of machine learning and deep learning approaches, including transfer learning, for EPI prediction.

Main Results:

• EPIs play a critical role in regulating gene expression.
• Numerous computational methods, including deep learning models, have been developed for predicting enhancers, promoters, and their interactions.
• Websites for accessing relevant datasets and tools are summarized.
• EPI identification methods are increasingly applied to study diseases like cancer.

Conclusions:

• Advances in computer technology, particularly deep learning and transfer learning, enable accurate prediction of EPIs from diverse genomic features.
• Deep learning models can directly predict EPIs from DNA sequences, reducing computational time for researchers.
• This review offers detailed research frameworks for scientists entering the field of enhancer-promoter interaction studies.