DILI risk screening at 0.989 AUROC.
96 public test compounds · identity, fingerprint and scaffold references excluded. The identity/fingerprint-only policy scores 0.952 on the same cases.
96 public test compounds · identity, fingerprint and scaffold references excluded. The identity/fingerprint-only policy scores 0.952 on the same cases.
Identity, fingerprint and test-scaffold references excluded. All 96 public test compounds scored.
The alternative reference policy on the same 96 public test compounds. All cases scored.
Both policies score all 96 public test compounds with the complete DILI system.
| Evaluation | AUROC | Coverage |
|---|---|---|
| Identities and identical fingerprints excluded | 0.952 | 96/96 cases scored; zero exact clinical self-anchors. |
| Additional test-scaffold exclusions | 0.989 | The same 96/96 cases scored. This is a second policy on the same known public test. |
| Output | Use |
|---|---|
| DILI risk score and class | Prioritize compounds for follow-up testing. |
| Mechanism context | Inspect the chemical and biological factors associated with a returned risk estimate. |
| Exposure and dose context | Evaluate the specified exposure scenario alongside the risk output. |
The AUROC measurements evaluate binary risk discrimination in the stated cohort. Mechanism attribution and exposure scenarios have their own validation scope. A frozen product and newly selected partner cases can test performance for your application.
Choose the properties, input conditions and scoring criteria for your application. A frozen system and partner-selected cases provide a direct product evaluation.