Guide · Drug discovery
In-silico vs traditional trials: what actually changes when your cohort is synthetic.
A comparison guide for clinical operations, translational science, and drug-discovery leaders evaluating where virtual patient populations fit alongside Phase I and Phase II human studies.
TL;DR
In-silico trials do not replace human studies. They compress the decisions made before a human study — dose, cohort, endpoint, and go/no-go — from years and eight figures to weeks and a compute budget.
1 · Definitions
Two very different instruments.
Traditional trial
A prospective study in recruited humans, governed by GCP and an IRB, designed to produce statistically defensible evidence of safety and efficacy for a regulator.
In-silico trial
A simulation of a candidate intervention on a synthetic patient population — a cohort sampled from mechanistic and data-driven priors — evaluated against the same endpoints a human study would use.
2 · Side-by-side
Where the two approaches diverge.
| Axis | Traditional Phase I/II | In-silico |
|---|---|---|
| Timeline per iteration | 12–36 months (protocol → IRB → recruit → dose → readout) | Hours to weeks (cohort synthesis → simulate → analyze) |
| Cost per candidate arm | USD 15M–70M for a Phase I/IIa program | Compute + validation: typically <2% of an equivalent human arm |
| Cohort size | 20–300 recruited humans, constrained by geography & consent | 10³–10⁶ virtual patients, sampled from validated priors |
| Rare subpopulations | Under-recruited; often excluded from primary analysis | Oversampled on demand (pediatric, hepatic impairment, comorbidities) |
| Failure cost | Sunk trial + patient exposure to a failing molecule | Discarded run; no human exposure |
| Regulatory role | Primary evidence for IND/NDA | Supportive: dose selection, adaptive design, MIDD packages (FDA) |
| What it cannot replace | — | Confirmatory safety and efficacy in humans |
3 · Cost & speed
The economics of asking the same question twice.
A conventional Phase I dose-finding study runs 6–12 months of active enrollment and burns most of its budget on site activation, monitoring, and per-patient costs that scale linearly with cohort size. Adding a subpopulation — pediatric, hepatic-impaired, a rare genotype — often means a second trial.
An in-silico arm collapses the same question onto a compute cluster. Cohort size stops being a cost driver; the marginal cost of the 1,000,001st virtual patient rounds to zero. Rare subgroups are oversampled by construction. And because you can re-run the study with a different dose, schedule, or endpoint in an afternoon, protocol design becomes iterative instead of fatal.
The realistic outcome isn't a trial you skip. It's a trial you enter already knowing the dose, the responders, and the failure modes — because you simulated them first.
4 · Where in-silico is the right tool
- 01Dose selection and dose-escalation strategy before first-in-human.
- 02Adaptive trial design — testing arm-adding and stopping rules under thousands of counterfactual populations.
- 03Rare-population extrapolation for pediatric, geriatric, and organ-impairment submissions.
- 04Failure triage — killing weak candidates in the discovery portfolio before they consume clinical budget.
- 05Comparator arms where a placebo or standard-of-care human arm is ethically or logistically constrained.
5 · Where it is not
In-silico evidence does not replace confirmatory human safety and efficacy data. Regulators — FDA under MIDD, EMA under its modelling-and-simulation guidance — accept in-silico evidence as supportive: it informs the study you run in humans, it does not substitute for it. A team that pitches virtual trials as an end-run around Phase III will lose credibility with both reviewers and their own clinical operations group.
6 · How Axonixx approaches this
Evaluating in-silico for your pipeline?
We run scoped pilots with drug-discovery and clinical-ops teams.