DEEPSEN: a convolutional neural network based method for super-enhancer prediction.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Super-enhancers (SEs) are critical regulatory elements driving cell identity genes.
• SEs are implicated in cancer oncogene expression and diseases like Alzheimer's.
• SEs offer potential for identifying oncogenes and disease mutation sites.

Purpose of the Study:

• To introduce DEEPSEN, a new computational method for predicting super-enhancers.
• To leverage convolutional neural networks for improved SE prediction accuracy.

Main Methods:

• DEEPSEN integrates 36 distinct features for SE prediction.
• The method employs convolutional neural networks (CNNs).
• Feature importance screening was performed to identify key predictive factors.

Main Results:

• DEEPSEN demonstrates superior performance compared to existing methods.
• The method enables accurate genome-wide super-enhancer prediction.
• Key features crucial for SE prediction were identified.

Conclusions:

• Convolutional neural networks significantly enhance super-enhancer prediction performance.
• DEEPSEN provides a powerful tool for SE identification and analysis.