Kill the wrong targets before they kill your pipeline.
Apoptosis Labs is the credit check for drug targets. Give it a target and a disease and it returns a calibrated GO or NO GO, plus the single cheapest, highest value experiment that would change the answer.
Mendelian randomisation replication in an independent biobank cohort
Resolves the weakest proof point, population generalisability, before any wet lab spend. Flips the verdict if the LDL-C effect fails to replicate.
- Time
- ~3 weeks
- Cost
- $12K
- Signal
- High
- Open Targets
- Ensembl
- UniProt
- ChEMBL
- RCSB PDB
- Literature
This is not rare
- 18 months
- spent validating a single target
- 86%
- of drugs entering trials never reach approval
- $500K
- lost per day spent on the wrong target
Most targets are a maybe, that is where we come in.
and exactly why
One serious liability stops a programme on its own. You get the reason, not just the verdict.
advance to the next gate
Graded evidence per axis, a calibrated confidence, and the citations behind every claim.
and the experiment that resolves it
We name the weakest link, then hand you the cheapest experiment that would settle it.
The pipeline
- 01
Resolve
Your target and disease are pinned to verified identifiers from live lookups.
- 02
Fan out
Specialist agents grade biology, safety, IP and more in parallel against evidence rubrics.
- 03
Score
One calibrated verdict. A single fatal finding can veto it, and thin evidence returns NOT SURE YET.
- 04
Route
The cheapest, fastest experiment that would change the answer, ranked by decision value.
The compounding loop
The more you put in, the better it works for you. Every result you send back re-scores the target, tightens the confidence and sharpens what we recommend next.
1 experiment in
- 01
Route
We pick the experiment most likely to change the decision, weighing signal against cost and time.
- 02
Report
You run it and send the result back. Positive, negative or inconclusive, it all counts.
- 03
Recalibrate
The target is scored again, and what you learned sharpens every later call on related biology.
Your experiments compound the more you use us.
Why this isn’t just asking a chatbot
We rank experiments, not just targets
The deliverable is the experiment itself, ranked by expected decision value net of cost and time. That layer is empty in every tool we compared against.
Calibrated, and willing to abstain
Confidence is calculated from evidence strength, reproducibility, directness and coverage. When the evidence is not strong enough to support a verdict, the system says so and hands back the experiment that would settle it.
Hard data, with the receipts
Evidence comes from live data sources rather than model memory. Resolved identifiers, data release versions and citations are attached to the underlying claims, including structural evidence, so a scientist can check the work.
Worked examples
Scored against decisions the field already made.
PCSK9 × hypercholesterolaemia
Human genetics, a dose responsive biomarker and approved precedent all point the same way. The clean positive control.
IL-23 vs IL-17A × Crohn's
Same disease, two targets, opposite outcomes, and the difference was visible before the trials confirmed it.
BACE1 × Alzheimer's
Target engagement was never the problem. A masking guard stops a strong biomarker story from carrying a failed clinical hypothesis.
Minutes to save years.
Run your first target × disease pair and see the verdict, the reasoning behind it, and the experiment that would change it.
New accounts join a waitlist until an admin approves them.