Revealing Hi-C subcompartments by imputing inter-chromosomal chromatin interactions.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Higher-order genome organization and its variability across cellular conditions are not well understood.
• High-coverage Hi-C data revealed distinct chromosomal subcompartments, but annotation is limited to specific cell lines due to data requirements.
• Comparing subcompartment patterns across diverse cell types is challenging with current methods.

Purpose of the Study:

• To develop a computational approach for inferring genome subcompartments using typical Hi-C datasets with moderate sequencing coverage.
• To enable practical comparison of subcompartment patterns across various cell types.
• To investigate the functional genomic features associated with conserved and cell-type specific subcompartments.

Main Methods:

• Developed SNIPER (Subcompartment iNference using Imputed Probabilistic ExpRessions), a computational method utilizing denoising autoencoders and multilayer perceptron classifiers.
• Applied SNIPER to moderate coverage Hi-C datasets.
• Compared SNIPER's performance against existing methods relying on epigenomic features.

Main Results:

• SNIPER accurately infers subcompartments from moderate coverage Hi-C data.
• SNIPER outperforms a previously established method in subcompartment annotation.
• Analysis of eight additional cell lines revealed distinct functional genomic patterns in conserved versus cell-type specific subcompartments.

Conclusions:

• SNIPER facilitates subcompartment identification without the need for high-coverage Hi-C data.
• The method provides valuable insights into the functional implications and mechanisms driving spatial genome organization variation.
• SNIPER is a significant advancement for studying genome architecture across different cellular contexts.