Compare: Overview · Feature Matrix · Cost · Benchmark · Alternatives
OverviewMetabarcoding software and pipelines compared.
Much of the day-to-day in metabarcoding isn't analysis. It's keeping the pieces in sync: matching formats between tools, pinning reference-database versions, redoing QC by hand. spcfy absorbs that upkeep. Pick your situation below, or jump straight to the numbers.
Which of these is you?
I have FASTQ files and just need an answer
You have sequencing files and a question. You don't want to build or run anything, and you've maybe wondered whether you could just hand the files to a chatbot.
See why a hosted platform beats the shortcut → ScientistI run analyses, but bioinformatics isn't my focus
You need results you can trust, but your time belongs to the study, not to keeping a pipeline alive.
Go to the full comparison → Lab bioinformaticianI keep pipelines alive and produce client reports
You maintain the analysis stack and turn it into reports for others. Automation for the routine work would free you for special requirements and statistics.
Go to the matrix, report angle → Method developer / PhDI'm building new pipeline logic, not running a fixed one
You're developing methods: a different clustering approach, a new taxonomic model, custom pipeline steps. You need full control of every step.
Probably not your tool, here's why →Oxford Nanopore support is live.
The long-read pipeline covers every marker the platform supports. Its first version starts from trimmed FASTQ files, not raw reads, so you enter the workflow one step in. The benchmark further on still reflects Illumina data only, because that is what the published study measured.
Talk to us about your ONT data →Four mechanisms most pipelines skip.
Hit selection that shows its work
Coverage-corrected scoring, alt hit kept on the row, not just the top BLAST hit.
No single database decides alone
For CO1, BOLD, NCBI and a MIDORI2-trained classifier are weighed against each other. Other markers run on their own reference.
One identity across every run
Results from different runs link by lookup, and spcfy re-annotates them with each major database release.
Clustering follows the marker
VSEARCH or SWARM clustering for most markers, UNOISE3 denoising for the 12S and 16S vertebrate markers.
spcfy
- Hosted, no compute or pipeline to maintain
- Fixed, versioned processing, comparable across labs and time
- Coverage-aware hit selection and cross-database consensus (CO1)
- Portable identity across runs and labs (SPIN), re-annotated with each database release
- Reports and querying built into the platform
QIIME2, nf-core, Apscale, Galaxy & co.
- You choose, install and maintain every step yourself
- Full control of every parameter, exactly what some projects need
- No built-in coverage-aware hit selection or consensus taxonomy
- Cross-run comparability is something you build yourself
- Compute, upkeep and reporting are on you
Everything you need to know.
Full feature matrix
Every capability, spcfy against MAP, mBRAVE, MetaWorks, QIIME2, nf-core/ampliseq and Apscale, row by row.
Open the matrix →What it actually costs
"Free" pipelines aren't zero-cost. Build, run, staff and compute time, converted into a fair comparison.
See the cost breakdown →Independent benchmark
Where spcfy lands in a non-peer-reviewed preprint, including where it doesn't win.
See the results →vs Nextflow, an LLM, or DIY
Operating models compared directly: orchestrating it yourself, prompting a chatbot, or running open source.
Compare alternatives →