ideasIdea
In silico trials to prioritise combinations, scored against later real trials
Simulate trials of drug combinations in populations of virtual patients to decide which real trials to run, and keep score of how often the simulations were right.
The number of possible combinations far exceeds trial capacity. Simulated trials using mechanistic and machine-learned models of virtual patient populations could rank combinations, but their predictive value is unknown. The proposal is a scored programme: simulations are registered with predicted effect sizes for combinations entering real phase 2 or 3 trials, and outcomes are compared as trials read out, building a public track record that determines how much weight simulation gets in portfolio decisions.
Hypothesis
Simulation rankings will correlate positively with real trial outcomes for at least some drug classes, allowing a measurable reduction in failed phase 3 combination trials when used to filter candidates.
Rationale
Regulators already accept in silico evidence for device testing and some pharmacokinetic questions; the missing element for efficacy is a track record, which only forward scoring can build.
What would test it
Register predictions for 50 ongoing combination trials; compare with outcomes as they read out over four years; publish the correlation and calibration.
Maturity
speculative
Who has to act
research
Cost to try
Medium ($1M to $50M)
Years to first evidence
4
Bottlenecks it attacks
- AI that is built but not validated or deployed · Thousands of cancer AI models are published; a handful are in clinical use, and fewer have shown they help patients.
- Too many combinations to test · There are thousands of possible drug pairs and sequences. Trials can test a few dozen a year.
- Trial design, endpoints and cost · A phase 3 trial takes years and hundreds of millions of dollars, and often answers a question that has already moved on.