iEnhancer-DS: Attention-based improved densenet for identifying enhancers and their strength

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Enhancers are critical DNA elements regulating gene expression through transcription factor binding.
• Accurate identification and strength prediction of enhancers are vital for understanding gene regulation.
• Current experimental methods are costly and time-consuming; existing computational methods have limitations in performance and encoding complexity.

Purpose of the Study:

• To develop an efficient and accurate computational method for enhancer identification and strength classification.
• To propose a deep learning-based multi-task framework, iEnhancer-DS, addressing limitations of current approaches.

Main Methods:

• Utilized one-hot encoding and nucleotide chemical properties (NCP) for DNA sequence feature embedding.
• Employed an improved DenseNet module for learning high-order implicit features.
• Integrated a self-attention mechanism for dynamic feature weighting and a multilayer perceptron (MLP) for prediction.

Main Results:

• iEnhancer-DS achieved state-of-the-art performance in both enhancer identification and strength prediction tasks.
• Demonstrated significant improvements in accuracy (ACC) and Matthews correlation coefficient (MCC) over existing methods.
• Enhancer identification saw ACC and MCC increases of 4.03% and 8.47%, respectively.
• Enhancer strength prediction showed ACC and MCC increases of 1.40% and 3.81%, respectively.
• Interpretable analysis using t-SNE visualized the model's mechanism of action.

Conclusions:

• iEnhancer-DS offers a powerful and accurate deep learning solution for enhancer identification and strength prediction.
• The framework provides a valuable tool for advancing research in gene regulation mechanisms.
• The study provides publicly available code and data for reproducibility and further development.