Safety & responsible use
In oncology, a confident wrong answer is worse than no answer. Provotics is designed around calibration, abstention, and transparency so its outputs stay honest and reviewable.
Provotics is a research and educational project. It is not a medical device, has not been cleared or approved by any regulator, and must not be used for clinical diagnosis, treatment decisions, or patient care. Its outputs are research hypotheses.
Predictions carry probabilities that mean what they say, so users can weigh a call instead of trusting it blindly.
Low-confidence and out-of-distribution profiles are flagged or abstained on, not forced into a label.
Each call exposes the driver genes behind it, so an expert can check it against known biology.
No identifiable patient data should ever be submitted. Inputs are expression profiles, not patient records. See our Privacy Policy.
Performance drops on fully independent, out-of-distribution cohorts, the expected generalization gap. Some rare anatomical sites have limited training data and weaker accuracy. Subtype and target predictions are hypotheses for expert review, not conclusions.
We publish these limits on purpose. A model that hides its failure modes is more dangerous than one that names them.
We're glad to talk through appropriate use, validation, and limitations before you apply.
Get in touch