ZipHiC: a novel Bayesian framework to identify enriched interactions and experimental biases in Hi-C data

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• Hi-C and related 3C-based methods generate crucial data for analyzing genome organization.
• Existing computational methods often fail to account for inherent dependencies within Hi-C data.
• Accurate analysis of Hi-C data is vital for understanding 3D genome structure and function.

Purpose of the Study:

• To introduce ZipHiC, a novel statistical method for detecting enriched contacts in Hi-C data.
• To address the limitations of existing methods by incorporating data dependencies.
• To provide a robust tool for analyzing Hi-C contact frequency matrices.

Main Methods:

• Developed ZipHiC, a Bayesian method utilizing a hidden Markov random field (HMRF) model.
• Employed Approximate Bayesian Computation (ABC) for efficient computation and interaction detection.
• Incorporated the Potts model to leverage neighbor information and account for biases.

Main Results:

• ZipHiC demonstrated superior performance over existing tools on both simulated and real Hi-C datasets.
• The method identified biases in Hi-C data, including those related to DNA accessibility and transposable elements.
• Analysis in Drosophila melanogaster revealed functional interactions between promoters and other genomic regions.

Conclusions:

• ZipHiC offers an advanced approach for Hi-C data analysis, improving the detection of significant genomic interactions.
• The method provides valuable insights into sources of bias affecting Hi-C and micro-C data.
• ZipHiC effectively corrects biases in micro-C datasets, enhancing their reliability.