EPIHC: Improving Enhancer-Promoter Interaction Prediction by Using Hybrid Features and Communicative Learning

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Enhancer-promoter interactions (EPIs) are crucial for gene regulation, cell differentiation, and understanding disease mechanisms.
• Experimental identification of EPIs is costly and time-consuming, necessitating efficient computational methods.

Purpose of the Study:

• To develop a novel deep learning model, EPIHC, for accurate prediction of Enhancer-Promoter Interactions (EPIs).
• To enhance the understanding of gene regulation by identifying functional relationships between enhancers and promoters.

Main Methods:

• Utilized deep neural networks with convolutional neural networks (CNNs) to extract sequence-derived features from enhancers and promoters.
• Incorporated a communicative learning module to capture inter-sequence information and integrated genomic features.
• Evaluated EPIHC on benchmark and chromosome-split datasets for prediction accuracy.

Main Results:

• EPIHC significantly outperformed existing state-of-the-art methods in predicting Enhancer-Promoter Interactions.
• The communicative learning module provided explicit, previously ignored information and offered a degree of explainability for EPIs.
• EPIHC demonstrated robust performance in cross-cell line predictions, suggesting generalizability.

Conclusions:

• EPIHC offers a powerful and efficient computational approach for identifying Enhancer-Promoter Interactions.
• The method's ability to integrate sequence and genomic features, coupled with communicative learning, advances the field of gene regulation prediction.
• Findings support the potential of EPIHC for broader applications in genomics and disease mechanism research.