EPIFBMC: A New Model for Enhancer-Promoter Interaction Prediction.

Area of Science:

• Genomics
• Epigenetics
• Computational Biology

Background:

• Enhancer-promoter interactions (EPIs) are critical epigenetic regulators of gene expression, influencing cellular identity and function.
• Understanding EPIs is essential for deciphering transcriptional regulatory networks in development, cell differentiation, and disease pathogenesis.

Purpose of the Study:

• To introduce EPIFBMC, a novel deep learning framework for accurate prediction of enhancer-promoter interactions.
• To leverage DNA sequence and genomic features for enhanced EPI prediction capabilities.

Main Methods:

• Developed EPIFBMC, a deep learning framework comprising Four-Encoding, Balanced Ensemble Subset Learning (BESL), and Multi-channel Network (MCANet) modules.
• Utilized DNA sequence and genomic features for training and prediction.
• Validated the model on multiple cell line datasets (HeLa, IMR90, NHEK) and cross-cell-line experiments (K562, GM12878, HUVEC).

Main Results:

• EPIFBMC demonstrated high accuracy in predicting enhancer-promoter interactions, outperforming existing state-of-the-art methods.
• The model achieved a balance between genomic feature richness and computational efficiency, significantly reducing training time.
• Ablation studies identified positional conservation and positional specificity score as key DNA sequence features for EPI prediction.

Conclusions:

• EPIFBMC provides a powerful and efficient tool for decoding gene regulatory networks through accurate EPI prediction.
• The framework holds significant potential for applications in developmental biology, disease mechanism research, and therapeutic target discovery.