BioinvestGPT ApS · Research use only
BVCT questions and answers
BVCT (BeatSoC Virtual Clinical Trial) is a patient-data-free platform by BioinvestGPT ApS that simulates a drug in a disease-matched virtual human and outputs a modeled clinical effect size against a named comparator, with the mechanistic rationale. Its outputs are for research use only. The answers below cover its inputs, outputs, track record, security and limits.
Last updated · Maintained by BioinvestGPT ApS
What does BVCT stand for?
BVCT stands for BeatSoC Virtual Clinical Trial. It is developed by BioinvestGPT ApS. SoC means standard of care; the name reflects the question BVCT models: whether, and by how much, a drug's effect size differs from that of the standard of care.
What is BVCT?
BeatSoC Virtual Clinical Trial (BVCT), developed by BioinvestGPT ApS, is a patient-data-free clinical AI infrastructure that simulates the mechanistic interaction between a drug and a disease-matched virtual human body, then outputs quantified clinical effect size against named comparators — including comparators never tested head-to-head in humans — while also providing the underlying mechanistic rationale. Research use only.
Is BVCT a clinical trial involving human participants?
No. BVCT is a computational platform that produces modeled estimates. Its virtual trials are distinct from clinical trials that enroll human participants, and its outputs are for research use.
Does BVCT need patient data?
No. BVCT is patient-data-free: a virtual trial is specified by the drug design and mechanism of action, the disease context, the comparator, the endpoint and the eligibility criteria. It draws on public scientific knowledge, which is distinct from private patient data.
What does BVCT output?
A quantified clinical effect size against a named comparator, stated in the result summary as a modeled hazard ratio (HR) range with bounds at one decimal place, together with the mechanistic rationale behind it and a GO/NO-GO interpretation. See the glossary for each measure.
What does "0.1 precision" mean?
Effect sizes are reported to one decimal place (for example, an illustrative predicted range of 0.4 < HR < 0.6) rather than as qualitative labels. That is display resolution, not measured prediction error; an error claim needs an endpoint-specific definition, an evaluation cohort and a reported error distribution.
Can BVCT compare a drug with a comparator it was never tested against?
Yes, as a modeled estimate. BVCT can output an effect size against a named comparator even where no head-to-head trial in humans exists. The result is a modeled comparison, not an observed one.
How accurate have BVCT predictions been?
As of 24 September 2026, BVCT's company-reported prospective track record is 95.6% accuracy (674/705 scored readouts), 89.7% positive predictive value (234/261) and 99.1% negative predictive value (440/444). A further 5 readouts could not be adjudicated. The scored predictions are dated calls of whether each trial would succeed, not effect-size estimates. The figures are company-reported, counted per scored readout; every record is on the public prediction dashboard. See the evidence page for the cohort definition.
Does a GO result guarantee clinical success?
No. A GO result is a modeled decision-support output under specified assumptions. It does not guarantee a trial outcome, regulatory approval or commercial success.
Where can I check a prediction?
Start with the public prediction dashboard at data.bioinvestgpt.com, then read the trial's outcome source (registry entry, sponsor report or publication). Check the timing, the endpoint match and the classification shown on the dashboard.
Does BVCT have a HIPAA-compliant architecture?
Yes. BVCT uses AES-256-GCM encryption, Argon2id password hashing, short-lived access tokens, re-authentication on reload and a hash-chained audit log. This describes technical safeguards, not a certification. See the security page.
Can BVCT results be used for decisions about individual patients?
No. BVCT works at population and trial level. Its outputs are research-use computational predictions for drug R&D and asset assessment: not medical advice, not for individual-patient diagnosis or treatment, and not treatment recommendations or regulatory approvals.