HAMPLE: deciphering TF-DNA binding mechanism in different cellular environments by characterizing higher-order...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Transcription factor (TF) binding to DNA is crucial for gene regulation but challenging to predict computationally.
• Understanding TF-DNA interactions requires characterizing complex nucleotide dependencies across cell types.

Purpose of the Study:

• To develop a computational framework, HAMPLE, for accurate prediction of transcription factor binding sites (TFBS).
• To characterize higher-order nucleotide dependencies and their role in TF-DNA binding mechanisms across diverse cell types.

Main Methods:

• HAMPLE utilizes a multi-task learning approach to predict TFBS in distinct cell types simultaneously.
• DNA sequences are represented using k-mer encoding, DNA shape, and histone modification features.
• A customized gate control and channel attention convolutional architecture captures cell-type-specific and shared motifs.

Main Results:

• HAMPLE significantly outperforms state-of-the-art methods in TFBS prediction, achieving superior auROC scores.
• Feature importance analysis confirms the predictive power and complementarity of k-mer encoding, DNA shape, and histone modification.
• Ablation studies and interpretable analysis validate the effectiveness of HAMPLE's architecture in capturing nucleotide dependencies.

Conclusions:

• HAMPLE provides an effective computational solution for predicting TFBS by integrating higher-order nucleotide dependencies.
• The framework enhances our understanding of TF-DNA binding mechanisms in various cellular contexts.
• The developed model and source code are publicly available for further research.