Proteome Focused Multi disease bacterial 16S and viral metagenomics collection

Protein-level data products for three published gut disease cohorts: inflammatory bowel disease (IBD; Norman et al. 2015, PRJEB7772), colorectal cancer (CRC; Hannigan et al. 2018, PRJNA389927), and graft-versus-host disease (GvHD; Thiele Orberg et al. 2024, PRJEB53547). This record builds on Zenodo Record 1, taking its per-ASV taxonomy tables, ASV sequence collections, and viral contig catalogues as input. Each cohort provides: Harmonized taxonomy — per-ASV taxonomy mapped to ProGenomes3 genome assemblies. ASVs are searched against a 16S subset of ProGenomes3 with MMseqs2, then assigned NCBI taxids and GCA accessions at species, genus, and family resolution via a three-layer assignment with enforced taxid uniqueness. Protein collections — bacterial proteins extracted from the ProGenomes3 representative-proteins reference for the target GCA set, viral proteins predicted from the contig catalogue with Prodigal in metagenomic mode, and the combined collection used for clustering. Protein-cluster matrices — bacterial and viral proteins clustered jointly in a single MMseqs2 run, so clusters may span both domains. Provided as a genome × cluster protein-count matrix and its binary presence/absence counterpart, together with cluster assignments and representative sequences. The joint cross-domain clustering makes these matrices directly usable for phage–host network inference. All processing parameters are documented in full in the accompanying README and pipeline_parameters.md, and the consumed Record 1 inputs are included in each cohort folder, so the outputs can be regenerated without downloading Record 1. This is Record 2micro in the series. Record 1 provides the upstream ASV inference, viral contigs, and abundance tables. Licensed CC-BY-4.0. Please cite Zenodo Record 1 and the original cohort studies alongside this dataset.