Romulus: robust multi-state identification of transcription factor binding sites from DNase-seq data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Transcription factor (TF) binding site prediction is crucial for understanding gene regulation but remains computationally challenging.
• Existing methods often struggle with accuracy, particularly for low-information-content motifs and diverse chromatin states.

Purpose of the Study:

• To develop a novel computational method, Romulus, for robust identification of individual TF binding sites.
• To improve the accuracy and robustness of TF binding site prediction by integrating genomic sequence and cell-type-specific experimental data.
• To introduce a quantitative measure of TF pioneer factor activity.

Main Methods:

• Romulus integrates genome sequence information with cell-type-specific DNase-seq data.
• The method reduces the number of free parameters, enhancing model robustness.
• Performance was assessed by comparing predictions against ChIP-seq profiles.

Main Results:

• Romulus significantly outperforms existing TF binding site prediction methods across multiple DNase-seq datasets.
• The method demonstrates superior performance for predicting binding sites of low-information-content motifs.
• Romulus can infer multiple binding modes for a single TF and introduces Binding in Closed Chromatin (BCC) as a measure of pioneer factor activity.

Conclusions:

• Romulus offers a robust and accurate computational approach for TF binding site prediction.
• The method enhances understanding of TF binding dynamics and pioneer factor function.
• Romulus is available as a free R package, facilitating broader research applications.