Screening services
Library triage run for you, rather than software you run yourself. Send a compound list and a target; get back a ranked, filtered, annotated shortlist and a written read on what to trust in it.
The pilot
Send 250 compounds against a target you already have assay data for. Public, blinded, or internal. I score them free and return the ranking within 48 hours, so you can grade it against results you already trust rather than take my word for anything.
If the ordering doesn't help on your chemistry, that is a genuinely useful answer and it costs you a CSV. I would rather find that out on 250 compounds than after an invoice.
What a project includes
- Ranked CSV. Predicted pKd, QED, LogP, TPSA, HBD/HBA, rotatable bonds, Lipinski compliance, PAINS/BRENK/NIH structural-alert flags, and your own series tags and IDs preserved.
- ChEMBL analog mapping. Nearest known compounds by Tanimoto similarity, with clinical-phase flags where an analog reached the clinic.
- Selectivity heatmap for multi-target runs, showing predicted margins across a family.
- Pocket viewer files with residue listings for the targets in scope.
- A short written read — what I would prioritise, and, more usefully, which parts of the output I would not rely on.
What the scoring can and cannot do
Measured on 516 protein–ligand complexes across 179 targets, every structure released after the model's training cutoff. Full workings, including a correction that moved these numbers, are on the validation page.
| Measure | Result | What it means |
|---|---|---|
| Top-quintile enrichment | 1.70× | The predicted top fifth of a diverse library holds ~1.7× as many potent compounds as a random fifth. |
| Top-decile enrichment | 2.23× | 51% hit rate against a 23% base rate. |
| Rank correlation | ρ = 0.39 | 0.63 on clean Kd labels; weaker on assay-dependent IC50. |
| Within one potency band | ρ = −0.01 | No ordering ability. It will not rank an SAR series, and I will not sell it as though it will. |
This is a filter for shortening a long, diverse list before you commit synthesis. It is not a replacement for docking, FEP, or an assay, and the absolute pKd should not be read as a predicted Ki.
Pricing
| Engagement | Scope | Price |
|---|---|---|
| Pilot | 250 compounds, one target, returned in 48 hours. | Free |
| Screen | Up to 5,000 compounds across one to three targets, with everything in the deliverables list above. | $1,500 |
| Generation campaign | Goal-directed molecule generation with REINVENT4 against your target and property profile, then re-scoring and selectivity profiling of the output. | Scope-dependent |
| Ongoing triage | Recurring screening as new target campaigns open. New libraries, not iteration within an existing series. | Let's talk |
On generation specifically: the reinforcement-learning loop optimises against the affinity model described above, so anything it produces inherits that model's limits. Treat generated structures as hypotheses to filter and inspect, not as a ranked shortlist. That is why it is priced per project rather than per compound.
Your compounds
Structures sent for scoring are held in memory, scored, returned, and dropped. They are not written to disk, not retained, and not shared. The specifics are published rather than summarised. If you need that in a signed agreement first, or would rather send a blinded list with identifiers stripped, both are fine.
Start with the pilot
Email a 250-compound CSV and the target you want it scored against, and say what you already know about the actives. Reply within 48 hours.
Zach Chiappini · VectaBind · validation · methods