Why genetic evidence matters for drug target validation

Most drug targets fail because a target that looks promising in a cell line or a small study can still turn out to be wrong — or unsafe — in humans, and that only becomes clear deep into clinical development. Fewer than one in eight drugs entering clinical trials reach approval (Wouters et al., JAMA 2020). By the time a wrong target reveals itself in the clinic, the cost is measured in years and hundreds of millions. The biggest driver of overall drug-development cost is failure itself (DiMasi et al., J Health Econ 2016) — which means the most powerful way to control cost is to choose targets that are more likely to be real before committing to them.

Yes. Targets with human genetic evidence are more than twice as likely to be approved (Nelson et al., Nature Genetics 2015), and the most recent analysis refines that to a 2.6-fold higher success rate from Phase 1 to approval (Minikel et al., Nature 2024). Human genetics works as an early signal because a genetic link between a gene and a disease in real human populations is direct evidence that modulating that gene may affect the disease — evidence that doesn't depend on a model system being faithful to human biology.

Yes — and this is the part that is least discussed. The advantage genetic evidence gives is not shrinking as the field matures; it is growing, because every new genetic finding adds evidence that wasn't there before (Minikel et al., Nature 2024). A practical consequence: a target assessed against the evidence 18 months ago may look different against today's evidence, and a target that looked weak then may be supported now. Evidence that is checked once and filed away goes stale. This is the direct justification for validating a target continuously rather than once.

Public databases and published studies tell you what has already been found in general — they do not tell you whether your specific target holds up, whether it replicates across more than one population, or whether the picture has changed since the last paper. A genetic signal that appears in one cohort does not always replicate in another; promising targets have looked strong in one dataset and vanished in the next. Turning published evidence into a decision about your target requires running it against real cohorts, replicating it separately across populations, and re-running it as variants are reclassified — slow, technical work most teams cannot spare infrastructure for.

CohortLayer tests your target or genetic hypothesis against large human cohorts and returns aggregate evidence — effect size, carrier counts, replication status, and phenotype and safety signals — never individual-level data. It replicates your hypothesis across more than one cohort separately, so what it reports has already survived a second test, and it re-runs continuously as the underlying evidence changes, flagging what moved. The individual data never leaves its secure cohort environment; only aggregate results are returned.