SeqEnhDL: sequence-based classification of cell type-specific enhancers using deep learning models.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Identifying cell type-specific regulatory elements is crucial for understanding gene regulation.
• Genome-wide computational identification of these elements presents a significant challenge.

Purpose of the Study:

• To develop a deep learning framework (SeqEnhDL) for accurate, genome-wide identification of cell type-specific enhancers.
• To leverage DNA sequence features for enhancer classification.

Main Methods:

• Utilized DNA sequences from ENCODE project's "strong enhancer" chromatin states across nine cell types.
• Employed positional k-mer fold changes as features for deep learning models (MLP, CNN, RNN).
• Compared SeqEnhDL against state-of-the-art enhancer classifiers like gkm-SVM and DanQ.

Main Results:

• SeqEnhDL models demonstrated superior performance in distinguishing cell type-specific enhancers from non-coding sequences.
• SeqEnhDL successfully discriminated enhancers from different cell types, a capability lacking in previous classifiers.
• Analysis confirmed that sequence features alone are sufficient for accurate enhancer and tissue-specificity identification.

Conclusions:

• SeqEnhDL provides a powerful deep learning approach for identifying cell type-specific enhancers based on sequence data.
• The framework advances computational methods for regulatory element discovery and understanding tissue-specific gene regulation.
• SeqEnhDL is publicly available, facilitating further research in the field.