OpusTaxa: A Unified Workflow for Taxonomic Profiling, Assembly, and Functional Analysis of Shotgun Metagenomes

Shotgun metagenomics has become a cornerstone of microbiome research, yet the complexity of existing workflows remains a major barrier for life scientists without dedicated bioinformatics support. Manual database setup, detailed sample sheet preparation, a nd management of software dependencies can make routine analysis difficult and time - consuming. Cross -study comparisons are further hampered by inconsistent processing pipelines, database versions, and profiling strategies, limiting reproducibility and the potential for large-scale meta-analyses. We present OpusTaxa, an open-source Snakemake workflow that provides end-to-end processing of short paired -end shotgun metagenomic data with minimal configuration. Users provide either FASTQ files or Sequence Read A rchive accessions; OpusTaxa automatically downloads required databases, performs quality control, removes host reads, and executes taxonomic profiling, metagenome assembly, and functional analysis. All analysis modules can be independently toggled, and per -sample outputs are automatically merged into harmonised, cross -sample tables ready for downstream exploration. Across two public datasets, we demonstrate how OpusTaxa can be used to compare consistency across complementary taxonomic profilers and to estim ate microbial load in addition to standard metagenomic workflows. Availability OpusTaxa is freely available at https://github.com/yenkaiC/OpusTaxa. Documentation, test data, and example configurations are included in the repository. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint

The global microbiome community has generated hundreds of thousands of metagenomes in recent decades . Y et differences in processing pipelines, database versions, and profiling strategies make cross -study comparisons difficult and often irreproducible. As bioinformatic

and reference databases evolve rapidly, even datasets from the past few years can yield new insights when re-analysed with updated tools. Similarly, as analysis pipelines differ in reference databases, underlying algorithms, and reporting conventions, results can be method-dependent [1], complicating comparisons across studies. At the same time, r unning multiple complementary tools on the same dataset can provide an internal sanity check, helping to distinguish stable biological signals from tool-specific artefacts. However, re-analysing past data with up -to-date databases and software versions is currently a time-consuming and complex process. Workflow management systems such as Snakemake [2] and Nextflow [3] allow users to simplify this process. While th ere are existing workflows for metagenome analyses (Table 1), they focus primarily on one of taxonomic profiling [4] or metagenome assembly [5-7], and often require users to manually install tools and download external databases [8, 9]. Additionally, few tools automate integration with data repositories to support reanalysis of public data [5]. To fill this gap, we developed OpusTaxa, an open -source Snakemake workflow designed to provide end -to-end metagenomic processing for life scientists with minimal bioinformatics expertise. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint Table 1: Feature comparison across publicly available workflows for shotgun metagenomic data analysis. Workflows were evaluated for workflow engine, container and sample sheet requirements, automated database provisioning, quality control and host removal strategies, taxonomic profiling tools, metagenome assembly, genome binning, functional and inference analyses, SRA integration, and generation of merged cross-sample output tables. Feature Workflow engine Container required Sample sheet required Auto database download Quality control Host removal Taxonomy Metagenome Assembly Genome Binning Functional Analyses Inference Analyses SRA integration Merged cross- sample tables Comments OpusTaxa Snakemake No No Yes fastp NoHuman Kraken 2 / MetaPhlAn / SingleM MetaSPAdes No HUMAnN / Resistance Gene Identifier / antiSMASH Microbial Load Predictor Yes Yes nf-core/taxprofiler Nextflow Yes Yes No fastp Bowtie2 Kraken 2 / MetaPhlAn / KrakenUniq / KMCP / MALT / DIAMOND / Centrifuge / Kaiju / mOTU / ganon No No No Yes nf-core/mag Nextflow Yes Optional No fastp / AdapterRemoval / Trimmomatic Bowtie2 GTDB-TK / CAT MEGAHIT / MetaSPAdes MetaBAT2 / MaxBin2 / CONCOCT / COMEBin / MetaBinner / SemiBin2 No Partial MetaFlowX Nextflow No Yes No fastp / Trimmomatic Bowtie2 Kraken2 / MetaPhlAn / GTDB- TK / HUMAnN MEGAHIT / MetaSPAdes MetaBAT2 / MaxBin2 / CONCOCT / COMEBin / MetaBinner / SemiBin2 / VAMB / binny VFDB / Resistance Gene Identifier / antiSMASH / eggNOG-mapper No Yes ATLAS Snakemake No No Partial BBTools BBMap GTDB-TK MEGAHIT / MetaSPAdes MetaBAT2 / MaxBin2 / VAMB / SemiBin2 / DAS Tool eggNOG-mapper / DRAM Yes Yes MAGNETO Snakemake No Yes No fastp + FastQ Screen Bowtie2 GTDB-TK / mOTU MEGAHIT MetaBAT2 eggNOG-mapper No No MetaWRAP Custom (Bash) No No No Trim Galore / BMTagger BMTagger Kraken 2 / Kraken MEGAHIT / MetaSPAdes MetaBAT2 / MaxBin2 / CONCOCT PROKKA No No MUFFIN Nextflow No No No fastp No sourmash with GTDB MetaSPAdes / Flye MetaBAT2 / MaxBin2 / CONCOCT EggNOG (KEGG pathways) No No Sunbeam Snakemake No No No Trimmomatic + Cutadapt + Komplexity bwa Kraken 2 (via extension) MEGAHIT (via extension) No No Yes No nf-core/funcscan Nextflow Yes Yes No SeqKit No No No AMPs (ampir / Macrel / AMPlify / HMMER) / ARGs (ABRicate / AMRFinderPlus / fARGene / RGI / DeepARG) / BGCs (antiSMASH / DeepBGC / GECCO / HMMer) No No funcscan is more a downstream contig- screening tool rather than an end- to-end metagenomics pipeline .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint Design philosophy and architecture OpusTaxa is designed to be a comprehensive and easy-to-use metagenomic profiling workflow manager for short-read paired -end shotgun metagenomic sequencing data, such as that generated by Illumina and MGI platforms (Figure 1). It is built on Snakemake [2] and offers both conda and containerisation for reproducible dependency management . The workflow is modular: every analysis tool can be toggled on or off via command (e.g. metaphlan=true, kraken2=false), allowing users to customise their analysis without modifying pipeline code. Figure 1: Pipeline overview of OpusTaxa. Modules are colour-coded by analysis category: data acquisition, quality control, host removal, taxonomic profiling, metagenome assembly, and functional analysis. A key design goal was to minimise the configuration burden. OpusTaxa automatically detects paired-end FASTQ files in the input directory, supporting common naming conventions (Illumina, SRA, dot -separated). All required databases for selected tools including those for NoHuman [10], MetaPhlAn [11], SingleM [12], Kraken2 [13], HUMAnN [14], antiSMASH [15], and CARD [16] are provisioned automatically on first run via checkpoints. Snakemake’s built-in job tracking ensures that previously completed samples are not reprocessed when the pipeline is re -run with additional data. For high -performance computing environments, OpusTaxa ships with the option of using a Snakemake profile for SLURM job submission, with configurable resource allocation per rule (CPU, memory, runtime). Default resource settings are tuned for typical metagenomic datasets but can be overridden in the configuration. HUMAnN3.9 antiSMASH FastQCFas tQC Kraken2 Resistance Gene Identifier (CARD) Pigz MetaPhIAn4 Local FastQ SRA ID SingleMFa stQ C SRA Tools Host Remov al (NoHuman) fastp Denov oassembly (m etaSPAd es) Taxonomic Analysis Functional Analysis OpusTaxa Microbial Load Predictor Inference Analysis .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint Input format and Sequence Read Archive integration OpusTaxa can either process local FASTQ files or download data directly from the Sequence Read Archive (SRA) [17] via fasterq‑dump. Downloaded reads are compressed with pigz [18] and automatically standardised from SRA to Illumina‑style filenames. Local and SRA‑derived samples can be mixed in a single run, making it straightforward to reanalyse published cohorts alongside new data. Quality Control and host-read removal Raw paired‑end reads are quality‑trimmed with fastp [19], which removes adapters, low‑quality bases, and short fragments. Quality reports are generated at three stages, raw reads, post‑trimming, and post‑host‑removal, using FastQC [20] and aggregated with MultiQC [21] to provide a clear view of data quality at each step. This three-stage reporting gives users a clear view of data quality at each processing step, enabling rapid identification of problematic samples. Human reads are removed using NoHuman [10], which employs a Kraken2-based classification against the Human Pangenome Reference Consortium (HPRC) r2 database. This approach offers an optimal host-read removal while maintaining computational efficiency, as pangenome databases better capture human genomic diversity than a single linear reference genome. Taxonomic profiling tools OpusTaxa currently integrates three complementary taxonomic profilers that differ in methodology and output. MetaPhlAn 4 [11] identifies species using clade -specific marker genes, producing relative abundance profiles. SingleM [12] takes a distinct approach, using single-copy marker genes built from the Genome Taxonomy Database to estimate both taxonomic composition and the prokaryotic fraction of the metagenome [22]. Kraken2 [13], paired with Bracken [23] for Bayesian abundance re -estimation, offers k -mer-based classification of each individual read , which includes archaea, bacteria, viruses, fungi, protozoa, and the human genome. These tools were selected to provide complementary taxonomic profiles based on distinct analytical strategies and reference databases, enabling users to compare results across fundamentally different approaches. Each profiler can be enabled independently, and OpusTaxa automatically merges per-sample profiles for each tool into cross-sample abundance tables for easy downstream analysis and comparison between samples. De-novo assembly and Functional Profiling When enabled, OpusTaxa performs per‑sample metagenome assembly with MetaSPAdes [24], retaining contigs and scaffolds and removing intermediate files to conserve storage. OpusTaxa supports both read‑based and contig‑based functional characterisation and automatically merges results across samples. For reads, HUMAnN 3.9 [14] quantifies gene families and metabolic pathways using UniRef90 and outputs are normalised to copies‑per‑million. As HUMAnN does not natively accept paired‑end input, it is run on forward reads only. For .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint contigs, Resistance Gene Identifier (CARD) [16] annotates antimicrobial resistance genes, while antiSMASH [15] detects biosynthetic gene clusters to infer secondary‑metabolite potential. To improve runtime and the quality of the output, contigs under 1000 base pairs are not included in functional analyses . OpusTaxa also estimates absolute microbial load from MetaPhlAn output profiles via Microbial Load Predictor [25], facilitating quantitative microbiome profiling. All per‑sample outputs are consolidated into harmonised, cross‑sample tables optimised for comparative analyses. OpusTaxa is developed in accordance with the FAIR Principles for Research Software (FAIR4RS) [26]. The source code is openly available on GitHub under MIT licence, with documentation, test data, and example configurations included in the repository. A versioned, citable release is archived on Zenodo (DOI: https://doi.org/10.5281/zenodo.19491844), providing a persistent identifier for reproducibility. Dependencies are fully specified via conda environment files, and all outputs conform to standard bioinformatics file formats to facilitate interoperability with downstream tools.

To demonstrate OpusTaxa’s utility, we first re -processed four publicly available human gut metagenome samples from the SRA. The five most abundant species identified in each sample by each of the three taxonomic profilers (Bracken, MetaPhlAn 4, and SingleM) were compared across samples ( Figure 2 ). Across all four samples, the three profilers recovered broadly concordant dominant community members, SingleM reported alternative scientific names in a small number of instances (e.g. Agathobacter faecis in place of Roseburia faecis), reflecting differences in taxonomy frameworks rather than major disagreement in biological interpretation. This concordance is notable because the three profilers operate with different databases and algorithms. Bracken, derived from Kraken2, uses k -mer-based classification of individual reads against a broad reference database; MetaPhlAn 4 infers species abundance from clade -specific marker genes; and SingleM uses universal single -copy marker genes derived from the Genome Taxonomy Da tabase. Agreement across these complementary approaches suggests that the dominant signals in these samples are robust to method choice We next applied OpusTaxa to a longitudinal dataset examining gut microbiome recovery following treatment with three broad -spectrum antibiotics (gentamicin, meropenem, and vancomycin) [27]. OpusTaxa downloaded the raw reads directly from the SRA, generated species-level abundance profiles with MetaPhlAn 4, and supplied these profiles to the Microbial Load Predictor (MLP). Shannon diversity calculated from MetaPhlAn 4 abundance profiles ( Figure 3a ) revealed a diverse and evenly distributed community at baseline that collapsed sharply following antibiotic administration, with diversity at its lowest on day 4. Partial recovery was observed by day 8, with near recovery at day 42, and complete recovery by day 180, reproducing reported patterns of post -antibiotic microbiome reconstitution [27]. Predicted microbial load ( Figure 3b) mirrored this trajectory, declining following antibiotic exposure and recovering by day 42 , showing similar results to a prior study [25]. Together, .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint these analyses show that OpusTaxa can recover biologically plausible and literature-consistent patterns from public metagenomic datasets while also extending interpretation through integrated quantitative modules such as MLP. Figure 2: Cross-tool taxonomic concordance across 4 control samples of human gut metagenome reanalysed from Duru et al., 2018 [28]. The five most abundant species identified per sample by Bracken, MetaPhlAn 4, and SingleM are shown. All samples were processed using OpusTaxa v0.3. 0 5 10 15 Phocaeicola doreiEscherichia coli Ruminococcus torques Bifidobacterium adolescentis Flavonifractor plautii Relative Abundance (%) ERR10818363 0 2 4 6 8 Bacteroides uniformisPhocaeicola vulgatus Bacteroides cellulosilyticus Agathobacter rectalisVescimonas coprocola Relative Abundance (%) ERR10818366 0 2 4 6 Faecalibacterium prausnitzii Roseburia faecisAlistipes putredinis Phocaeicola vulgatusBacteroides uniformis Relative Abundance (%) ERR10818373 0.0 2.5 5.0 7.5 10.0 Roseburia faecis Agathobacter rectalisPhocaeicola vulgatusBacteroides uniformis Alistipes putredinis Relative Abundance (%) ERR10818516

Bracken MetaPhlAn SingleM Top Overlapping Species by Sample and Tool .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint Figure 3: Recovery of gut microbiota following broad-spectrum antibiotic exposure, reanalysed from Palleja et al., 2018). Subjects were treated with gentamicin, meropenem, and vancomycin. (a) Shannon diversity index calculated from MetaPhlAn 4 species-level abundance profiles across sampling timepoints. (b) Predicted microbial load (cells per gram) estimated by Microbial Load Predictor across the same timepoints. All samples were processed using OpusTaxa v0.3. Statistical comparisons between day 0 and each subsequent timepoint (days 4, 8, 42, and 180) were performed using the Wilcoxon rank-sum test.

OpusTaxa provides an end -to-end Snakemake workflow for shotgun metagenomic analysis that combines taxonomic profiling, metagenome assembly, and functional analysis in a single, minimally configured framework. Its design priorities : automatic database provisioning, no sample sheet requirement, conda-only dependency management, and built-in SRA integration; distinguish it from existing workflows and lower the barrier for life scientists who lack dedicated bioinformatics support. OpusTaxa is in active development with further taxonomic profilers planned (for example mOTU [29] and others in consideration ), as well as virome profiling and additional functional profiling tools. Currently, OpusTaxa deliberately omits genome binning , and does not perform taxonomic profiling of assembled contigs , though OpusTaxa’s MetaSPAdes outputs can be used as input for these tools if desired To our knowledge, OpusTaxa is the first published workflow to integrate SingleM and the Microbial Load Predictor directly within an end-to-end shotgun metagenomic processing framework. In summary, OpusTaxa enables rapid, scalable analysis of either public or local metagenomic data with up-to-date analysis tools and databases in as few as five short commands from installation to results.

We thank colleagues for helpful discussions that informed development of OpusTaxa and users for testing. Large language models were utilized to assist with grammar and spelling corrections, as well as for code documentation. Funding This study was financially supported by grants from the National Health and Medical Research Council of Australia (F.R.: APP2017404). 0.37 0.00014 0.057 0.00016 0 1 2 3 4 0 4 8 42 180Days after treatment Shannon Diversity 0.59 0.015 0.8 0.0014 10.25 10.50 10.75 11.00 11.25 0 4 8 42 180Days after the treatment Predicted load (log10) a b .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint CRediT authorship contribution statement Yen-Kai Chen: Conceptualization, Software, Formal analysis, Validation, Visualization, Writing original draft. Clarice M. Harker: Validation, Writing review & editing. Cong M. Pham: Validation, Writing review & editing. Luke Grundy: Supervision, Writing review & editing. Hannah R. Wardill: Supervision, Writing review & editing. Michael J. Roach: Conceptualization, Software, Methodology, Supervision, Writing review & editing. Feargal J. Ryan: Conceptualization, Methodology, Supervision, Funding acquisition, Writing review & editing.

1. Pelto, J., et al., Elementary methods provide more replicable results in microbial di5erential abundance analysis. Briefings in Bioinformatics, 2025. 26(2): p. bbaf130. 2. Mölder, F ., et al., Sustainable data analysis with Snakemake. F1000Res. 2021; 10: 33. 2026. 3. Di Tommaso, P ., et al., Nextflow enables reproducible computational workflows. Nature biotechnology, 2017. 35(4): p. 316-319. 4. Stamouli, S., et al., nf-core/taxprofiler: Highly parallelised and flexible pipeline for metagenomic taxonomic classification and profiling. BiorXiv, 2023: p. 2023.10. 20.563221. 5. Kieser, S., et al., ATLAS: a Snakemake workflow for assembly, annotation, and genomic binning of metagenome sequence data. BMC bioinformatics, 2020. 21(1): p. 257. 6. Churcheward, B., et al., MAGNETO: an automated workflow for genome-resolved metagenomics. Msystems, 2022. 7(4): p. e00432-22. 7. Krakau, S., et al., nf-core/mag: a best-practice pipeline for metagenome hybrid assembly and binning. NAR genomics and bioinformatics, 2022. 4(1): p. lqac007. 8. Uritskiy, G.V ., J. DiRuggiero, and J. Taylor, MetaWRAP—a flexible pipeline for genome-resolved metagenomic data analysis. Microbiome, 2018. 6(1): p. 158. 9. Xia, Y ., et al., MetaflowX: a scalable and resource-e5icient workflow for multi- strategy metagenomic analysis. Nucleic Acids Research, 2025. 53(18): p. gkaf954. 10. Hall, M.B. and L.J. Coin, Pangenome databases improve host removal and mycobacteria classification from clinical metagenomic data. GigaScience, 2024. 13: p. giae010. 11. Blanco-Míguez, A., et al., Extending and improving metagenomic taxonomic profiling with uncharacterized species using MetaPhlAn 4. Nature biotechnology, 2023. 41(11): p. 1633-1644. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint 12. Woodcroft, B.J., et al., Comprehensive taxonomic identification of microbial species in metagenomic data using SingleM and Sandpiper. Nature biotechnology, 2025: p. 1-6. 13. Wood, D.E., J. Lu, and B. Langmead, Improved metagenomic analysis with Kraken 2. Genome biology, 2019. 20(1): p. 257. 14. Beghini, F ., et al., Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3. elife, 2021. 10: p. e65088. 15. Blin, K., et al., antiSMASH 8.0: extended gene cluster detection capabilities and analyses of chemistry, enzymology, and regulation. Nucleic acids research, 2025. 53(W1): p. W32-W38. 16. Alcock, B.P ., et al., CARD 2023: expanded curation, support for machine learning, and resistome prediction at the Comprehensive Antibiotic Resistance Database. Nucleic acids research, 2023. 51(D1): p. D690-D699. 17. Katz, K., et al., The Sequence Read Archive: a decade more of explosive growth. Nucleic acids research, 2022. 50(D1): p. D387-D390. 18. Adler, M., pigz: A parallel implementation of gzip for modern multi-processor, multi-core machines. Jet Propulsion Laboratory, 2015. 19. Chen, S., fastp 1.0: An ultra-fast all-round tool for FASTQ data quality control and preprocessing. Imeta, 2025. 4(5): p. e70078. 20. Andrews, S., FastQC: a quality control tool for high throughput sequence data. 2010. 2017. 21. Ewels, P ., et al., MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics, 2016. 32(19): p. 3047-3048. 22. Eisenhofer, R., A. Alberdi, and B.J. Woodcroft, Large-scale estimation of bacterial and archaeal DNA prevalence in metagenomes reveals biome-specific patterns. bioRxiv, 2024: p. 2024.05. 16.594470. 23. Lu, J., et al., Bracken: estimating species abundance in metagenomics data. PeerJ Computer Science, 2017. 3: p. e104. 24. Nurk, S., et al., metaSPAdes: a new versatile metagenomic assembler. Genome research, 2017. 27(5): p. 824-834. 25. Nishijima, S., et al., Fecal microbial load is a major determinant of gut microbiome variation and a confounder for disease associations. Cell, 2025. 188(1): p. 222-236. e15. 26. Chue Hong, N.P ., et al., FAIR principles for research software (FAIR4RS principles). Zenodo, 2022. 27. Palleja, A., et al., Recovery of gut microbiota of healthy adults following antibiotic exposure. Nature microbiology, 2018. 3(11): p. 1255-1265. 28. Duru, I.C., et al., Metagenome-assembled microbial genomes from Parkinson’s disease fecal samples. Scientific Reports, 2024. 14(1): p. 18906. 29. Ruscheweyh, H.-J., et al., Cultivation-independent genomes greatly expand taxonomic-profiling capabilities of mOTUs across various environments. Microbiome, 2022. 10(1): p. 212. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 19, 2026. ; https://doi.org/10.64898/2026.04.15.718825doi: bioRxiv preprint