LEAP: Long sequence enhancer activity analysis and prediction framework.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Enhancer activity is crucial for gene expression regulation.
• Predicting enhancer activity is challenging due to the complex relationship between DNA sequences and function, particularly concerning sequence scale.
• Existing computational methods often overlook the importance of longer DNA sequence information.

Purpose of the Study:

• To develop an interpretable framework for predicting enhancer activity.
• To investigate the impact of sequence scale on enhancer activity prediction.
• To analyze key sequence motifs and their functional importance through multi-angle experiments and cross-cell validation.

Main Methods:

• Developed LEAP (Learn Enhancer Activity and Predict), an interpretable framework utilizing the Performer model.
• Integrated gene information within a 1001bp range of enhancers to capture longer-scale sequence information.
• Performed multi-angle experiments and cross-cell validation to analyze critical motifs.

Main Results:

• LEAP outperforms state-of-the-art methods in enhancer activity prediction, achieving high PCC (0.794) and low MSE (0.406).
• The framework effectively captures longer-scale DNA sequence information for improved prediction accuracy.
• Contribution score visualization identified heterogeneous functional importance across critical motifs.

Conclusions:

• LEAP offers a novel perspective for predicting and analyzing enhancer activity.
• The study highlights the significance of DNA sequence length and specific motifs in enhancer function.
• LEAP provides insights into enhancer grammar and the importance of long-sequence enhancers in gene regulation.