Computational modeling of chromatin accessibility identified important epigenomic regulators

Area of Science:

• Genomics
• Epigenetics
• Molecular Biology

Background:

• Chromatin accessibility is crucial for gene transcription.
• Transcription factors (TFs) and histone modifications (HMs) are known regulators of chromatin accessibility.
• Quantitative studies on the predictive power of TFs and HMs are limited.

Purpose of the Study:

• To develop and apply a model to predict chromatin accessibility using DNA sequence, TF binding, and HM signals.
• To quantify the relative contributions of DNA sequence, TFs, and HMs to chromatin accessibility.
• To investigate cell-line specific and shared regulatory mechanisms.

Main Methods:

• Construction of a two-layer predictive model integrating DNA sequence, TF binding, and HM data.
• Application of the model to human cell lines GM12878 and HepG2.
• Analysis of feature importance for prediction accuracy.

Main Results:

• DNA sequence data showed limited predictive power for chromatin accessibility.
• Both TF binding and HM signals accurately predicted chromatin accessibility.
• HM prediction was consistent across cell lines, driven by five core HM types.
• TF models identified both cell-line specific and generic informative TFs.
• Combined TF and HM models did not improve prediction accuracy, suggesting redundant information.
• Both models accurately distinguished proximal versus distal transcription start site accessibility.

Conclusions:

• TF binding and HM signals are key determinants of chromatin accessibility, surpassing DNA sequence information.
• Histone modifications exhibit largely shared regulatory roles across cell lines.
• Transcription factors contribute through both cell-line specific and general mechanisms.
• TF and HM data provide largely overlapping information regarding chromatin accessibility.