Machine Learning Methods for Exploring Sequence Determinants of 3D Genome Organization.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Higher eukaryotic cells organize chromosomes within the nucleus.
• Advances in whole-genome mapping reveal multiscale 3D genome organization linked to genome functions.
• DNA sequence determinants of 3D genome organization are not well understood.

Purpose of the Study:

• To review computational approaches for identifying DNA sequence elements critical for 3D genome organization.
• To highlight the role of machine learning in connecting DNA sequence to 3D genome architecture.
• To identify future research directions for improved predictive models.

Main Methods:

• Review of recent computational and machine learning-based methods.
• Analysis of approaches predicting 3D genome organization from DNA sequence.
• Discussion of the integration of multi-omic data and model interpretability.

Main Results:

• Machine learning methods are rapidly advancing the prediction of 3D genome organization from DNA sequence.
• These methods facilitate understanding sequence features at various scales of genome architecture.
• Progress has been made in developing predictive models for sequence elements in 3D genome organization.

Conclusions:

• Computational and machine learning approaches are crucial for deciphering sequence determinants of 3D genome organization.
• Future research should focus on integrating multi-omic data and enhancing model interpretability.
• This will deepen our understanding of gene regulation through the 3D genome structure.