ColdDDI
ColdDDI: Evaluating Knowledge Utilization in Cold-Start Drug-Drug Interaction Prediction
A diagnostic benchmark with 565,731 DDI pairs from DrugBank evaluating model reliance on mechanistic evidence in cold-start settings.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info coldddi
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
2 sourced entriesEvery number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
Built from DrugBank 5.1.13, it contains 1,900 drugs and 565,731 positive pairs, stratified by drug novelty (S0-S2) and mechanism (PK/PD, Type A/B). It uses masking and sensitivity metrics to assess if models utilize provided knowledge graph evidence rather than just accessing it.