Compare: Overview · Feature Matrix · Cost · Benchmark · Alternatives
Compare directlyChoose the setup that fits how your lab works.
Operating models, orchestrating it yourself, prompting a chatbot, or running open source. What to check before you commit follows below.
How you compute matters as much as what runs.
Not a feature-for-feature comparison, these describe different ways of getting from FASTQ to a result, not competing tools.
| Attribute | spcfy | Galaxy | Seqera / nf-core | Managed service | LLM API |
|---|---|---|---|---|---|
| Direct cost | subscription + per sample | none | usage (compute) | high, per run | very high |
| Process / staff effort | low | medium to high | high (setup) | low | high |
| Deterministic & reproducible | ● | ● | ● | by contract | – probabilistic |
| Data stays under your control | ● DE / GDPR | ● | ● | by contract | – to vendor |
| Turnaround | hours | queue-dependent | your compute | days per project | varies |
Reproducible only if you run the engine.
nf-core/ampliseq is a gold standard for reproducibility, but it's reproducibility of your orchestration, on your compute, with databases you choose and maintain per marker. spcfy gives fixed reproducibility without you operating the engine, plus hit-selection and cross-run identity a stock ampliseq run doesn't include.
spcfy fits when
- You have no compute cluster, or don't want to run one
- You want the same fixed processing every run
- You need cross-run comparability (SPIN)
Nextflow/nf-core fits when
- You have HPC or cloud and the skills to drive it
- You need full control of every parameter
- Setup, upkeep and the compute bill are yours
Not either/or in practice: many labs run nf-core for method control and use spcfy as the fixed, comparable reference, or the other way round.
You don't have to become a bioinformatician.
If you have sequencing files and just need an answer, a hosted platform runs the whole path on fixed, documented steps, and hands you results you can repeat and defend.
A fixed pipeline
- Same steps, same cutoffs, every run
- Run it twice, get the same answer
- Data processed in the EU, files stay yours
An LLM as the pipeline
- Probabilistic: each run can choose different steps
- Two runs of the same data can disagree
- Raw FASTQ leaves your control, typically to a US-hosted model
Most labs would prefer not to send raw FASTQ into a US-hosted general LLM, and many want to run a model of their own. spcfAI accounts for that: it runs on spcfy's own model, or it connects through the API to a Claude or ChatGPT model you control. Either way it reads finished results instead of producing them.
Free of charge but not free of effort.
Galaxy (OBITools eDNA workflow) and Apscale (open-source) cost no license. They cost your time instead: tool choices, QC, version bookkeeping, and the compute underneath.
spcfy fits when
- You want a finished standard, not a toolbox to configure
- Cross-lab comparability matters (SPIN, fixed filters)
Running it yourself fits when
- License budget is the binding constraint
- You need full control and freedom to tune every threshold
| Cost line | Self-hosted open source | spcfy |
|---|---|---|
| License | none | subscription + per sample |
| Pipeline build + upkeep | your hours, ongoing | included |
| Cross-lab comparability | build it yourself | SPIN + fixed filters |
Where staff time is effectively free, the open-source column wins outright. Where it isn't, the fee and the hidden hours converge fast, see the cost breakdown.
What to measure in a vendor pilot.
Websites don't settle a procurement decision. Run identical data through spcfy and one self-hosted reference pipeline, and measure the right things: on a hosted platform you never pay for wall-clock or RAM, so those are not what to compare. Total cost of ownership and capability are.
Can the pipeline even do it?
- Coverage-aware hit selection, and whether the alt hit is reported at all
- Consensus taxonomy across more than one reference (spcfy: CO1 today)
- An identity that survives across runs and labs, and whether old results get re-annotated when a reference database is updated
- Which parameters you can set yourself, and which are fixed for everyone
Governance
- Data location, retention and deletion policy
- Where the AI stream runs, and on whose infrastructure
- What you can export, and whether a re-run reproduces it
- Which intermediate files you get, and which stay with the provider
Mostly in development: one piece is live today, three are on the roadmap.
Publication-ready output and where it stands.
For state institutes and researchers with FAIR requirements, output has to travel: standardized, documented, and submittable to public repositories without re-formatting. One part of that is in place today, the rest is on the roadmap, and it is worth knowing which is which before you plan a publication around it.
Full provenance on every export
Pipeline version, reference-database version, primers and QC stay attached to the result, and the tool versions used are reported back to you.
Darwin Core export
Output formatted for the biodiversity-data exchange standard, so it moves without re-formatting.
DOI-ready metadata
A metadata set complete enough to register a DOI against the dataset.
Direct repository submission
GBIF, ENA and BOLD, plus custom targets. Export and submit is a manual step until then.
Where that leaves you.
Five pages of comparison come down to one question: do you want to operate the analysis, or do you want the result? If you need control over every parameter and have the people to use it, an open-source stack is the better fit, and this comparison says so in several places. If you would prefer to spend that time on the study itself, that is what spcfy is built for.
You have seen the benchmark where spcfy loses, the capabilities still on the roadmap, and the two parameters you can actually set. The part a website cannot settle is how your own data behaves. Send us a dataset and find out.
Benchmark figures come from a non-peer-reviewed preprint (Prosser et al. 2026) by a competing pipeline's authors and rest on a single CO1 dataset. Feature claims about third-party tools are compiled to the best of current knowledge and may change, check each tool's own documentation when it matters. Roadmap items are marked as such throughout; self-hosting and customer-owned storage are not standard tiers but can be set up on request. spcfAI is live in alpha. Scope across all five pages is environmental biodiversity data only, no human genetics.
Compare: Overview · Feature Matrix · Cost · Benchmark · Alternatives