Skip to content

HiFi-MAG Pipeline

Long-read metagenomics within Galaxy — a Snakemake-based workflow (originally PacBio's pb-metagenomics-tools) for recovering high-quality Metagenome-Assembled Genomes (MAGs) from PacBio HiFi reads, packaged into Galaxy by the Schatz lab alongside new profilers and ties to the BioDIGS soil-metagenomics dataset. Schatz lab's contribution to Galaxy's metagenomics stack.

  • Authors: Tyler Collins, Alex Ostrovsky, Michael Schatz (JHU — Schatz lab)
  • Galaxy tool / platform: TBD — verify from the talk slides (likely on usegalaxy.org tool panel)
  • Source: PacificBiosciences/pb-metagenomics-tools (upstream Snakemake) · Galaxy wrappers TBD
  • Documentation / training: HiFi-MAG-Pipeline tutorial
  • Status at GBCC2025: oral talk on long-read metagenomics within Galaxy

Talk at GBCC2025

  • Speakers: Tyler Collins, Alex Ostrovsky, Michael Schatz
  • Day / session: Day 2 (Tue Jun 24, 2025) — Oral Session 2B
  • Talk title: "Broadening Metagenomic Horizons in Galaxy: Introducing the HiFi-MAG Pipeline, New Profilers, and Insights from BioDIGS"
  • Slides: TBD
  • Video: GBCC2025 playlist
  • Abstract / schedule: GBCC2025 program

What it does

The pipeline takes HiFi metagenomic assemblies and runs a "completeness-aware" binning strategy: contigs >500kb are screened with CheckM2, those >93% complete are pulled directly into the final MAG set, the rest are binned with MetaBat2 and SemiBin2, merged via DAS_Tool, QC'd with CheckM2 again, and taxonomically assigned with GTDB-Tk. The GBCC talk extends this with new taxonomic profilers and showcases application to BioDIGS (a citizen-science soil microbiome project run out of JHU).

How it works

Core idea. A Snakemake workflow that recovers MAGs from pre-assembled PacBio HiFi metagenomes by short-circuiting the binning step for already-complete long contigs and only running the expensive binning ensemble on what remains.

Inputs / outputs. Per sample: SAMPLE.fasta (HiFi reads) and SAMPLE.contigs.fasta (assembly produced upstream by hifiasm-meta, metaMDBG, or metaFlye — the pipeline recommends these but does not run them itself). Outputs: MAG FASTA files, summary tables joining CheckM2 quality with GTDB-Tk taxonomy, and plots of completeness, contamination, genome sizes, and read-mapping stats.

Key innovation. Completeness-aware binning (v2 strategy). Contigs >500 kb are evaluated independently with CheckM2; any that exceed 93% completeness are pulled directly into the final MAG set without going through binning, sidestepping the failure mode where binners fragment a high-quality single-contig genome. Remaining contigs are pooled and binned with MetaBat2 + SemiBin2, merged by DAS_Tool, re-QC'd by CheckM2, and taxonomically assigned by GTDB-Tk.

Parameters / API surface worth knowing. - Contig screening threshold — min_length (default 500 kb) for the direct-promotion path. - Completeness gate — 93% CheckM2 completeness for direct promotion. - Filter thresholds — completeness ≥70%, contamination <10%, contigs per bin <20 (all configurable). - Binners — MetaBat2 + SemiBin2 (long-read-tuned), merged by DAS_Tool.

Canonical example. From the tutorial: provide HiFi reads and contigs per sample, run snakemake --use-conda, and the workflow emits a MAG set with a summary table linking each bin to its CheckM2 quality and GTDB-Tk lineage. The GBCC talk's headline application is BioDIGS soil samples, where HiFi MAGs resolve strains a short-read pipeline would collapse.

Where it fits in the corpus

  • AACR 2026: axis = bioinfo-tools (microbiome / cancer-microbiome); long-read MAGs are the gold standard for strain-resolved cancer-microbiome work
  • mia (entry) — complementary roles: HiFi-MAG = upstream pipeline producing MAGs; mia = downstream R/Bioconductor analysis container for the resulting community profiles
  • EuroBioC 2025: mia + miaverse on the analysis side
  • Nextflow Summit 2026: parallel to nf-core/mag

Notes

PacBio HiFi long-read MAGs increasingly outperform short-read assemblies for strain-level resolution — relevant for cancer-microbiome studies that need to discriminate closely related taxa. The Schatz-lab Galaxy wrapping makes the pipeline accessible to biologists who don't want to drive Snakemake directly.