MINTsC learns multi-way chromatin interactions from single cell high throughput chromatin conformation data.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Single-cell high-throughput chromatin conformation (scHi-C) captures genome 3D organization but remains under-utilized.
• Current analyses focus on pairwise interactions and chromosomal structures, overlooking multi-way interactions.

Purpose of the Study:

• To introduce MINTsC, a novel computational method for learning multi-way chromatin interactions from scHi-C data.
• To address the under-utilization of scHi-C datasets by enabling the analysis of complex, multi-way genomic interactions.

Main Methods:

• MINTsC utilizes a dirichlet-multinomial spline model.
• It aggregates pairwise interactions across cells and summarizes them using order statistics of pairwise test statistics.
• The method provides well-calibrated p-values for false discovery rate control.

Main Results:

• Evaluation on cell line and complex tissue scHi-C datasets supports MINTsC-inferred multi-way interactions.
• Application to human prefrontal cortex data revealed multi-way interactions with implications for gene regulation by enhancers.
• MINTsC-inferred interactions show potential for molecular QTL and epistasis studies by reducing multiple-testing burden.

Conclusions:

• MINTsC effectively learns and analyzes multi-way chromatin interactions from scHi-C data.
• The method enhances the utility of scHi-C datasets for understanding genome 3D organization and gene regulation.
• MINTsC offers a powerful tool for genetic association studies, particularly for epistatic effects.