The model is sequence, ligand, sequence, and the steps below follow that order: the first target, then the compound, then the second target. Add more targets to rank a panel.
Targets are named by UniProt accession. Gene symbols are many-to-many against accessions and the failure is silent, so the accession is what the model is given and what the results show.
The one compound put to every target.
Up to 12 targets per run. Every pair is scored, so n targets is n(n−1)/2 comparisons.
Two targets never needs an email. A longer panel does, and the report is emailed back.
Unlocks the extra target box below ↓
Results go to that address, and we use it to say when the model changes.
Strength is the larger of the two returned probabilities. Every answer is labelled with its band and that band's measured accuracy.
| Strength at or above | Share kept | Accuracy |
|---|---|---|
| 0.50, answer everything | 100.0% | 0.750 |
| 0.60 | 74.4% | 0.815 |
| 0.70 | 52.0% | 0.877 |
| 0.80 | 35.9% | 0.933 |
| 0.90 | 22.1% | 0.966 |
A near-tie is a near-tie: where the two measured values sit within half a log the model is right 0.573 of the time.
The model answers which of two targets binds a compound more tightly. It is not a safety or toxicity screen: breadth across families in this data partly records how many assay panels a compound went through, not how promiscuous it is.