Normalizing Metagenomic Hi-C Data and Detecting Spurious Contacts Using Zero-Inflated Negative Binomial Regression

HiCBin: binning metagenomic contigs and recovering metagenome-assembled genomes using Hi-C contact maps

“…Apart from a state-of-the-art normalization method HiCzin [27], several relatively simple metagenomic Hi-C normalization methods have been developed. Beitel et al [19] divided raw Hi-C contacts by the product of the length of two contigs.…”

“…As the wastewater microbiome community sample was processed by the proprietary Hi-C preparation kit, the restriction enzymes utilized in the experiment were not specified, resulting in the lack of information on the number of restriction sites for contigs. Therefore, HiCzin_LC mode was employed for normalization [27]. Moreover, compared to the human gut dataset, the wastewater dataset was much more complicated with 752,580 assembled contigs longer than 1000 bp.…”

“…Apart from chromosomal contacts of interest, two types of experimental biases have been reported to have substantial effects on raw metagenomic Hi-C contact maps, rendering the normalization of Hi-C contact maps essential for the binning [27]. Therefore, we employed HiCzin, a state-of-the-art metagenomic Hi-C normalization method, to remove the influences of biases [27]. We denote the intra-species contacts as the count of Hi-C interactions between contigs from the same species and refer to the non-zero intra-species contacts as valid contacts.…”

“…We recently put forward three explicit experimental biases for raw metagenomic Hi-C contacts: the number of restriction sites, contig length, and contig coverage, and demonstrated that normalization methods in publicly available Hi-C-based binning pipelines cannot correct all explicit biases [27]. Those pipelines implementing normalization by the number of restriction sites cannot even be carried out when the restriction enzymes employed in Hi-C experiments are not specified.…”

“…To solve these problems, we develop HiCBin, a new open-source metagenomic Hi-Cbased binning pipeline, to recover high-quality MAGs. We employ HiCzin [27], a novel normalization method designed for metagenomic Hi-C contact maps, to remove the experimental biases. HiCBin also detects and removes the spurious inter-species contacts for the first time among all Hi-C-based binning pipelines.…”