Revisiting Assessment of Computational Methods for Hi-C Data Analysis

Abstract

The performances of algorithms for Hi-C data preprocessing, the identification of topologically associating domains, and the detection of chromatin interactions and promoter–enhancer interactions have been mostly evaluated using semi-quantitative or synthetic data approaches, without utilizing the most recent methods, since 2017. In this study, we comprehensively evaluated 24 popular state-of-the-art methods for the complete end-to-end pipeline of Hi-C data analysis, using manually curated or experimentally validated benchmark datasets, including a CRISPR dataset for promoter–enhancer interaction validation. Our results indicate that, although no single method exhibited superior performance in all situations, HiC-Pro, DomainCaller, and Fit-Hi-C2 showed relatively balanced performances of most evaluation metrics for preprocessing, topologically associating domain identification, and chromatin interaction/promoter–enhancer interaction detection, respectively. The comprehensive comparison presented in this manuscript provides a reference for researchers to choose Hi-C analysis tools that best suit their needs.