Deployment
Deployment, data handling and scope
Two ways, and the second is the main one
Installed on the institution's network
A pre-configured machine, graphics hardware included, delivered and installed inside the hospital or private network, kept current by us. It runs entirely within those walls. This is the option for teams bound by rules that keep data inside the country, or working with patient data that cannot cross a border.
Hosted by us
The same system on our infrastructure, with nothing to install and no hardware to procure. Suited to work that is not required to keep its data inside a particular country, and the quicker way to run an evaluation.
A gene symbol and a variant. That is the whole input.
No variant call file, no raw sequencing, no name, no medical record number, no date of birth. Biofacta is given something of the form KCNJ11 c.601G>A p.(Arg201His) and nothing that identifies a person. On an on-premises installation nothing leaves the network at all, including that.
What goes in
- A gene symbol, checked against known biology before anything runs.
- One coding variant in standard notation.
- Optionally, the interaction partners the team wants considered.
- Optionally, the institution's own list of what its pharmacy stocks.
What never goes in
- Variant call files or raw sequencing reads.
- Any identifier: name, medical record number, date of birth, address.
- Free-text clinical notes or phenotype descriptions.
- Imaging, pathology or any other patient record.
Operational shape
Updates
Model versions, the molecule corpus and the availability data are delivered as versioned bundles the institution applies on its own schedule. Every result records which versions produced it, so applying an update never silently changes a past finding.
Network access
The appliance is designed to run without outbound internet access. Updates can be delivered on media where policy requires it.
Hardware
Supplied with the machine rather than requisitioned from the institution, so a deployment does not wait on a graphics-hardware procurement cycle.
Audit
Every run is recorded with its inputs, its model versions, its caveats and its timestamp, and the record exports with the result rather than staying in the system.
The findings belong to the team
What a team discovers using Biofacta belongs to that team. We are the instrument, not a claimant on the result. There is no clause by which a finding, a target or a hypothesis becomes ours, and no requirement to route a result through us to publish it.
We are the instrument, not a claimant on the result.
Research use only, by design
Biofacta is not a medical device and makes no clinical claim. It produces mechanism and hypotheses for laboratory triage, and a clinician or scientist decides what follows. That is a scope rather than a limitation: the outputs are built to be read by a research team and defended to a review committee, which is a different artifact from a diagnostic report.
The delivery shape is already accepted
In 2022 King Faisal Specialist Hospital and Research Centre in Saudi Arabia deployed a commercial variant-interpretation platform on-premises, as an API, specifically so that genomic data would not leave the Kingdom.
That is somebody else's contract with somebody else's vendor, cited here as market evidence that a sovereign buyer will pay for delivery inside the perimeter. It is not a relationship of ours, and we have none to report yet.
What an evaluation needs
Four things, and the fourth is optional.
- 01A protein the team already cares about, with a sequence and ideally a solved structure.
- 02A missense variant. Loss-of-function findings are out of scope, and we would rather say that now.
- 03A laboratory that could test a hypothesis if one comes back worth testing.
- 04Optionally, the pharmacy's stock list, which is what turns a candidate into something obtainable this month.