Revisiting Assessment of Computational Methods for Hi-C Data Analysis.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Existing Hi-C data analysis tools lack comprehensive evaluation using recent, validated datasets.
• Previous performance assessments often relied on synthetic or semi-quantitative data, limiting real-world applicability.

Purpose of the Study:

• To comprehensively evaluate 24 state-of-the-art methods for the entire Hi-C data analysis pipeline.
• To provide a benchmark for selecting appropriate Hi-C analysis tools based on validated performance.

Main Methods:

• Utilized manually curated and experimentally validated benchmark datasets for evaluation.
• Included a CRISPR dataset for specific validation of promoter-enhancer interactions.
• Assessed methods for Hi-C data preprocessing, TAD identification, and interaction detection.

Main Results:

• No single method excelled across all evaluation metrics and analysis stages.
• HiC-Pro demonstrated balanced performance in Hi-C data preprocessing.
• DomainCaller and Fit-Hi-C2 showed strong, balanced performance in TAD identification and interaction detection, respectively.

Conclusions:

• The study offers a critical reference for researchers navigating Hi-C analysis tool selection.
• HiC-Pro, DomainCaller, and Fit-Hi-C2 are recommended for specific stages of Hi-C data analysis.
• Validated benchmarking is crucial for advancing Hi-C data analysis methodologies.