One standard across many labs.
Data from many researchers can only be pooled when the methods behind it match. Methods captures each lab's method, in the researcher's own words, and checks it against your standard before the data moves.
The data looks alike
Two labs measure the same thing and report the same numbers.
The methods do not
One drops disagreeing replicates by a rule, one by eye. One checked its reference gene, one never did.
Nobody can tell
None of that is in the data or the repository. The method is the missing metadata.
Three things, none of which needs a lab's data.
Program standard
What every lab in your program must state about its methods. Written once, with you. Every lab's session asks about it.
Fit report
One per lab: what it stated, what it never stated, what does not match your standard. Before ingestion, not after.
Divergence view
One step across every lab: who states a rule, which rules differ, who never stated one. It tells you which harmonization problem to solve first.
One lab's qPCR analysis, checked against an example standard.
| Your standard requires | The lab said | Fit |
|---|---|---|
| Instrument software version | version Not stated | Not stated |
| A replicate exclusion rule | "I drop the odd one. I go by eye." | Not stated |
| Reference gene stability checked | GAPDH · "I have never checked." | Not stated |
| Efficiency measured | Assumed 100% | Does not match |
| Error bars from biological replicates | Technical replicates | Does not match |
| Fold change as 2−ΔΔCt | 2−ΔΔCt vs GAPDH | Fits |
Three values to ask for, two differences to see before pooling, one that already fits. No data changed hands.
One step, twelve labs.
| Rule as stated | Labs | What you can do |
|---|---|---|
| Exclude when Ct standard deviation exceeds 0.5 | 4 | Make it the program rule |
| Exclude when Ct standard deviation exceeds 0.3 | 2 | Reconcile with the 0.5 group |
| Exclude the replicate farthest from the median | 1 | Flag before pooling |
| Not stated · "by eye" | 5 | Ask each lab for the rule |
Same capability. Different consent.
Your own research organization
- You set the standard; every internal lab's sessions use it
- You see every fit report and the divergence view
- Each record is still signed off by the researcher who made it
- A custom model on the data you already own, with the same honest survey and phased path a single lab gets
A funder with many external labs
- Labs opt in, one by one
- You see fit reports and the divergence view, never a lab's record unless the lab shares it
- The lab stays the author and the owner
- A model on pooled data only from labs that opted in for that purpose, and never from a lab's record
We never hold the data
- Raw data never enters Methods, at any tier, for any customer
- Records export to your repository with the method attached
- We compare methods; we never touch the numbers
Where this stands
The single-lab product runs with the early cohort this autumn. The program standard, fit report and divergence view are designed on top of what those sessions already produce. Next: one transformation view across all your labs. Where each lab's work sits on the ladder, the wins each has put to work, and what is being reused across labs, with each lab sharing only what it agrees to.
What we are looking for
One organization to write the first program standard with us. The question we would ask you: what would you need to know about a lab's methods before you would trust its data in a pool?