Know your pipeline is still validated after every change.
Aver benchmarks clinical pipelines against public reference materials, flags any change that moves the numbers, and signs the evidence.
Works with the stack you already run
Nextflow
nf-core
Snakemake
Docker
Conda
Python
PyPI
GitHub
The real app, running on a sample lab.
Open it when you're ready. Scroll inside it freely, and leave any time.
Inside the demo: click a node on the trend line, press Ctrl K to search, or Esc to leave.
From a changed dependency to signed evidence.
Four checkpoints, run automatically on every change.
- 01
Fingerprint
Every tool, container digest, parameter and reference, hashed.
- 02
Benchmark
Against GIAB or SEQC2 truth sets with hap.py, vcfeval or som.py.
- 03
Detect drift
A change re-runs the benchmark and measures it against your baseline.
- 04
Sign evidence
A deterministic, signed pack, ready for your quality system.
Pipelines change even when nobody touches them.
Aver records each change the moment it happens, with its measured effect.
| Detected | What changed | Severity | Measured impact |
|---|---|---|---|
| Day 1 | bwa-mem2 container rebuilt upstream | Documentation | Not re-run |
| Day 14 | samtools 1.21 → 1.22 | Review required | ±0.00 pp |
| Day 29 | GATK 4.5.0.0 → 4.6.0.0 | Revalidation required | −0.87 pp |
| Day 41 | multiqc pinned to :latest | Critical | Approvers alerted |
One real run, recorded end to end.
nf-core/sarek 3.10.0 on GIAB HG002 v4.2.1, chromosome 20, compared with hap.py.
Five new modules, one validation record.
Each feeds the same drift detection, audit history and evidence packs.
Run QC monitoringNew
Trend each run's metrics against control limits your lab sets. Flags name the rule that fired.
DRAGEN™ pipelinesNew
Snapshot the configuration from each run's replay file. Vendor upgrades show up as drift.
SomaticNew
Benchmark against the SEQC2 HCC1395/HCC1395BL reference pair.
Structural variantsNew
SVs of 50bp and up, scored with truvari on GIAB HG002 Tier 1.
PharmacogenomicsNew
Diplotypes checked against the CDC GeT-RM consensus, 34 genes.
DRAGEN™ is a trademark of Illumina, Inc. Aver is not affiliated with or endorsed by Illumina.
Change impact summary
| Change | Impact | Disposition |
|---|---|---|
| samtools 1.21 → 1.22 | ±0.00 pp | Acknowledged |
| bwa-mem2 digest rebuilt | Not re-run | Recorded |
| Scheduled verification | +0.01 pp | Within baseline |
sha256:3a667de0f…Evidence an inspector can check, not just read.
For labs working under CAP, CLIA, ISO 15189 or IVDR expectations. Aver generates the record; your lab makes the decisions.
- DeterministicSame period, byte-identical document.
- Signed with re-authenticationIdentity, role and time bound to one version.
- Hash-chained historyAny edit breaks the chain.
See a real number before you pay anything.
Open source CLI
Evaluate Aver locally, for research or before involving anyone else.
- Local benchmarking and reports
- Dependency fingerprinting
- Snapshot diff
Lab
For laboratories running accredited clinical testing that need evidence on demand.
- Continuous monitoring and alerts
- Run QC trending
- Hosted history and evidence packs
- Up to three pipelines
Enterprise
For IVDR or GxP teams with signature workflows in a quality system.
- Unlimited pipelines
- E-signatures and checklist mapping
- Audit export, SSO, priority support
Annual billing available. Purchase orders supported on Enterprise.
Questions laboratories ask first.
Short answers. The documentation has the detail.
What is Aver?
Aver is a continuous validation platform for clinical bioinformatics pipelines. It benchmarks a pipeline against public reference materials such as GIAB and SEQC2, detects any change that moves the numbers against the validated baseline, and produces signed, deterministic evidence that an inspector can verify.
How do I revalidate a clinical pipeline after a software update?
Aver fingerprints every tool, container digest, parameter and reference in the pipeline. When one changes, for example GATK 4.5.0.0 to 4.6.0.0, it re-runs the benchmark with hap.py, vcfeval or som.py, measures the result against your baseline in percentage points, and records the outcome for review and sign-off.
What is pipeline drift?
Pipeline drift is any change to a validated pipeline's behaviour after validation, including changes nobody made on purpose: a container rebuilt upstream, a tag such as :latest, or a dependency version bump. Aver grades each change with versioned rules as informational, documentation, review required, revalidation required or critical.
Does Aver need patient data?
No. Aver validates pipelines only against public reference materials, and it has no field that can hold a patient identifier or clinical sample. Its integrations are read-only, and the benchmarks can run entirely inside your network with the open-source CLI or the lab-hosted Runner.
Which pipelines and workflow managers does Aver support?
Aver resolves Nextflow pipelines, including nf-core pipelines such as sarek, and Snakemake workflows, with containers pinned by digest and conda environments. For DRAGEN it snapshots the vendor configuration from each run's replay file, so a vendor upgrade shows up as drift.
Which reference materials and benchmarking tools does Aver use?
Germline small variants are compared with GIAB truth sets using hap.py or vcfeval. Somatic variants use the SEQC2 HCC1395/HCC1395BL pair with som.py. Structural variants of 50 bp and up use GIAB HG002 Tier 1 with truvari, and pharmacogenomics uses the CDC GeT-RM consensus for 34 genes.
Does Aver approve a pipeline for clinical use?
No. Aver produces validation evidence for laboratories operating under CAP, CLIA, ISO 15189 or IVDR expectations. Decisions about whether a pipeline is fit for clinical use remain with the laboratory and its inspectors.
How can an inspector check Aver's evidence?
Evidence packs are deterministic, so the same period produces a byte-identical document. Each pack is signed after re-authentication, which binds identity, role and time to one version, and the audit history is hash-chained, so any edit breaks the chain and shows up on verification.
How much does Aver cost?
The command-line tool is free and open source. The Lab plan starts at $500 per month for up to three pipelines, with continuous monitoring, run QC trending and hosted evidence packs. Enterprise starts at $2,000 per month, with unlimited pipelines, e-signatures, checklist mapping and SSO.
How do I try Aver?
Install the open-source CLI with pip install aver-bio, then run aver init, aver resolve and aver benchmark. It runs fully locally and needs no account. You can push results to the platform later for continuous monitoring.
Run your first benchmark in minutes.
Fully local. No account needed. Push results to the platform when you want continuous monitoring.
# install and run locally pip install aver-bio aver init --preset nf-core/sarek --clinical aver resolve aver benchmark # HG002 · GRCh38 · chr20