Tracking shared microbial strains in Galaxy: SameStr tools and workflow now available!

Two microbiome samples can contain the same bacterial species and still carry different strains, and those strains can affect their host in very different ways. Strain-level resolution makes it possible to answer questions that species-level profiling cannot: after a fecal microbiota transplantation (FMT), which donor strains established themselves in the patient, and which of the patient’s own strains persisted?

SameStr (Podlesny et al., 2022) was designed for exactly this. It builds single-nucleotide variant profiles from species-specific MetaPhlAn marker genes and compares them across samples, detecting shared strains even when they are not the dominant strain in a sample. Until now, using it meant either installing software, building reference databases and working on the command line or using Nextflow Workflow Manager. With this release, the full analysis is available in Galaxy, and no command-line experience is needed.

From bachelor thesis to community resource

Xènia Morera Martínez, a bachelor’s student, carried out this work as her bachelor thesis in the Bioinformatics Group at the University of Freiburg, under the supervision of Saim Momin. Within a single project, she wrapped a complete tool suite and its Data Manager, built and published a peer-reviewed workflow, validated it against the reference implementation, and wrote a GTN tutorial. That covers the full cycle of contributing to Galaxy, and we are proud of her achievement.

The workflow

The new IWC workflow SameStr-based shared strain detection from paired-end metagenomic data takes a collection of paired-end shotgun metagenomic reads and runs the complete analysis:

1. Pre-processing: quality trimming and host-read removal with KneadData (Trimmomatic and Bowtie2)
2. Taxonomic profiling and marker alignment with MetaPhlAn4
3. Strain detection with the SameStr suite: Convert, Merge, Filter, Stats, Compare and Summarize
4. Results: a strain co-occurrence table, a per-clade strain events table (shared_strain, other_strain, same_clade) and per-sample taxon counts

Reference databases and the key overlap and similarity thresholds are exposed as workflow inputs, so every analysis is transparent and easy to adapt. The workflow is tested, versioned and citable (doi:10.5281/zenodo.23841287).

Eight new tools for the community

To build the workflow, all eight SameStr tools were wrapped and contributed to the Intergalactic Utilities Commission. They include SameStr Extract for reference genomes and a Data Manager that builds SameStr reference databases from MetaPhlAn or mOTUs markers. Galaxy administrators can therefore provide ready-to-use databases, and users simply select one from a dropdown. Each tool can also be used on its own, outside the workflow.

Validated against the reference implementation

The Galaxy workflow was benchmarked against the official SameStr Nextflow pipeline on five FMT-treated recurrent Clostridioides difficile infection (rCDI) cases from the FRICKE cohort used in the original SameStr study (ENA project PRJEB39023). Each case includes a donor sample, a patient sample taken before FMT and one or two patient samples taken after FMT, giving 16 samples in total. For every case, two comparisons were evaluated:

• Engraftment: strains shared between the donor and the post-FMT sample
• Persistence: strains shared between the pre-FMT and post-FMT samples

Both pipelines used MetaPhlAn4 with the same marker database (mpa_vJan25_CHOCOPhlAnSGB_202503) and identical SameStr settings: a minimum alignment identity of 90 %, a marker truncation of 20 nucleotides, and a shared strain called when two samples overlap on at least 5,000 marker positions with at least 99.9 % similarity. The only intended difference was pre-processing. This workflow follows the original SameStr publication (KneadData with Trimmomatic and Bowtie2 against the human genome), while the Nextflow pipeline uses BBDuk and Kraken2. Both of those tools are also available in Galaxy and could be swapped in.

• Pre-processing: KneadData was slightly more permissive, keeping on average 1.46 % more reads per sample (0.33-2.44 %) after quality trimming and host removal.

• High agreement on shared strains: across the 12 engraftment and persistence comparisons, Nextflow reported 186 shared strains and Galaxy 183. Of these, 176 were called by both. Only 10 were called by Nextflow alone and 7 by Galaxy alone, so about 91 % of all shared-strain calls agree. Six of the 12 comparisons matched strain for strain.

• Why the remaining 17 calls differ: these were borderline cases rather than disagreements in the method. In 15 of them, both pipelines analysed the same clade, but the slightly different set of reads changed a handful of nucleotide positions. That was just enough to move the similarity score across the 99.9 % cutoff in one pipeline and not the other, and every miss was within 0.0013 of the threshold. One clade fell just short of the 5,000-position overlap in Galaxy (4,698 positions), and one was detected in only a single sample by Nextflow, so it could not be compared there. Neither pipeline was systematically stricter. Comparing the pre-processed reads directly confirmed that the differences come from the pre-processing step and not from the SameStr strain detection.

• Biology reproduced: applying the original publication’s genus-level analysis to the Galaxy results recovered its key finding, that the genera which most often persist in healthy guts (such as Faecalibacterium, Eggerthella and Adlercreutzia) are also those which most often engraft after FMT.

Learn it with the Galaxy Training Network

A new GTN tutorial, Shared strain detection with SameStr, walks you through the analysis using real FMT data. It explains why strain-level analysis matters, shows you how to run the complete workflow, and goes through each SameStr tool step by step, including what every key parameter does and when to be careful about changing it.

Acknowledgements

We warmly thank Daniel Podlesny and the EMBL team behind SameStr and its Nextflow implementation, for developing these tools and for supporting their integration into Galaxy. We thank de.NBI (German Network for Bioinformatics Infrastructure) for providing the virtual machine used to benchmark both pipelines and to develop the blueprint of the Galaxy implementation. We are also grateful to Wolfgang Maier, Matthias Bernt and Marius van den Beek for their careful review of the wrappers and the IWC workflow.

Get started

• Run the workflow: import it from the IWC workflow page into usegalaxy.eu, where all tools and a current MetaPhlAn-based SameStr database are already installed.
• Follow the tutorial on the Galaxy Training Network.
• Coming next: support for mOTUs as an alternative profiler.

Questions or feedback? Join us on the Galaxy Help Forum or the Galaxy Matrix channels.