Systematic evaluation of statistical methods for identifying looping interactions in 5C data
Chromosome-Conformation-Capture-Carbon-Copy (5C) is a molecular technology based on proximity ligation that enables high-resolution and high-coverage inquiry of long-range chromatin looping interactions. Computational pipelines for analyzing 5C data involve a series of inter-dependent normalization procedures and statistical methods that markedly influence downstream biological results. A detailed analysis of the trade-offs inherent to all stages of 5C analysis has not been reported, but is essential for understanding the biological basis of looping. Here, we provide a comparative assessment of method performance at each step in the 5C analysis pipeline, including sequencing depth and library complexity correction, bias mitigation, spatial noise reduction, distance-dependent expected and variance estimation, modeling, and loop detection. We present a detailed discussion of methodological advantages/disadvantages at each step and provide a full suite of algorithms, lib5C, to allow investigators to test the range of approaches on their own high-resolution 5C data. Principles learned from our comparative analyses will have broad impact on many other forms of Chromosome-Conformation-Capture-based data, including Hi-C, 4C, and Capture-C.