Unveiling chromatin dynamics with virtual epigenome.

Area of Science:

• Genomics and Epigenetics
• Computational Biology
• Systems Biology

Background:

• Chromatin's 3D organization is crucial for gene regulation and cellular functions, with the epigenome playing a central role.
• Hi-C techniques reveal chromatin interactions but are costly and complex, limiting widespread application.
• Current predictive models for chromatin interactions often use limited ChIP-seq data, impacting accuracy and generalizability.

Purpose of the Study:

• To develop a computational approach, EpiVerse, for accurate prediction of chromatin interactions using imputed epigenetic signals.
• To enhance model interpretability by integrating chromatin state prediction within a multitask learning framework.
• To provide a comprehensive view of chromatin structure across 39 human tissues and enable in silico perturbation experiments.

Main Methods:

• Leveraged imputed epigenetic signals and advanced deep learning techniques.
• Employed a multitask learning framework incorporating chromatin state prediction.
• Applied the model to predict Hi-C contact maps across 39 human tissues.

Main Results:

• EpiVerse significantly improves the accuracy of cross-cell-type Hi-C prediction compared to existing methods.
• The model enhances interpretability by predicting chromatin states alongside Hi-C contacts.
• Generated comprehensive Hi-C contact maps for 39 human tissues, revealing complex chromatin structure-gene regulation relationships.

Conclusions:

• EpiVerse offers a powerful and accurate computational tool for predicting chromatin interactions and understanding 3D genome organization.
• The approach facilitates novel in silico experiments to explore chromatin architecture under specific conditions.
• EpiVerse advances the study of epigenetics and its role in gene regulation across diverse human tissues.