PyMEGABASE: Predicting Cell-Type-Specific Structural Annotations of Chromosomes Using the Epigenome.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Interphase genome folding in eukaryotes is classified into compartments and subcompartments using Hi-C data.
• These genomic structures exhibit distinct epigenomic features and vary by cell type.
• Understanding the link between genome structure and epigenome is crucial for cellular function.

Purpose of the Study:

• To develop a computational model, PyMEGABASE (PYMB), that predicts genome (sub)compartment annotations from local epigenomic data.
• To explore the relationship between epigenomic marks and genome structure.
• To provide a user-friendly and robust tool for analyzing genome organization.

Main Methods:

• A maximum-entropy-based neural network model (PYMB) was developed.
• The model predicts (sub)compartment annotations using epigenomic data, such as ChIP-Seq for histone modifications.
• PYMB was trained on human cell data and tested on human and mouse cell types.

Main Results:

• PYMB accurately predicts genome (sub)compartments from epigenomic data alone, without requiring Hi-C experiments.
• The model demonstrates cross-species transferability, predicting mouse compartments from human-trained data.
• PYMB provides interpretable predictions, highlighting the importance of specific epigenomic marks for subcompartment identity.

Conclusions:

• PYMB offers a novel method for annotating genome structure using epigenomic data, complementing traditional Hi-C approaches.
• The model's predictions are valuable for studying genome organization, cell identity, and gene expression.
• PYMB's interpretability and ability to generate 3D genome models with OpenMiChroM advance the field of genome architecture research.