In October 2026, TRAIRC published a signed summary of its Phase 0 scientific calibration using public computational drug-discovery data.
The calibration reported reproducible exact-identity findings in two conditions: a known LIT-PCBA PKM2 overlap case and an adapted OpenADMET–ExpansionRx control.
The original OpenADMET continuous-label version could not be evaluated under the production workflow, so a binary adaptation was used. That deviation was not authorized by the original precommitment, and it is disclosed in a separately signed public correction.
These results are an internal production calibration on public data. They are not independent external reproduction, prospective blind validation, or proof of general model correctness.