Benchmarking of 4C-seq pipelines based on real and simulated data

Area of Science:

• Genomics and Epigenetics
• Computational Biology
• Bioinformatics

Background:

• Chromosome conformation capture combined with high-throughput sequencing (4C-seq) is a powerful next-generation sequencing technique for epigenetic analysis.
• 4C-seq data are complex, prone to biases, and lack comprehensive benchmarking and realistic simulation tools.
• Existing specialized programs for 4C-seq analysis have not undergone extensive, unbiased evaluation.

Purpose of the Study:

• To conduct an unbiased, extensive benchmarking of 4C-seq analysis algorithms.
• To develop a novel simulation software for generating realistic 4C-seq data.
• To compare the performance of various 4C-seq algorithms under different data characteristics.

Main Methods:

• Benchmarking of 66 4C-seq samples from 20 datasets.
• Development of Basic4CSim, a novel 4C-seq simulation software.
• Comparison of algorithms on 50 simulated datasets with varying sample sizes.
• Inclusion of novel differential pipeline versions of single-sample algorithms.

Main Results:

• Identified significant differences in precision, recall, interaction structure, and runtime among 4C-seq algorithms.
• Observed general trends in algorithm performance across different scenarios.
• Found no single tool optimal for both near-cis and far-cis interactions, or for both single-sample and differential analyses.
• r3Cseq, peakC, and FourCSeq showed high precision for near-cis, while fourSig excelled in far-cis analyses.

Conclusions:

• Algorithm choice significantly impacts 4C-seq analysis results.
• 4C-seq simulations are valuable for developing improved analysis algorithms.
• Basic4CSim provides a resource for future algorithm development and validation.
• Further research is needed to optimize tools for specific 4C-seq analysis needs.