The view that genetic evidence should be the highest-weighted input in target prioritization has become something close to orthodoxy in drug discovery over the last decade. The logic is compelling and the clinical validation data supports it. Targets with strong human genetic evidence for disease involvement have a materially higher success rate in clinical development. But the orthodoxy is sometimes applied as a binary filter rather than a weighted tier, and that application produces its own failure mode.
Why Genetic Evidence Earns Its Position
The causal argument for genetic evidence rests on Mendelian randomization. Inherited variants are assigned at conception, before disease onset, and are largely unaffected by the disease process or its treatment. When a coding variant that disrupts gene function is associated with disease risk in a population cohort, the direction of causation is clear in a way that expression data, protein abundance, or literature evidence cannot match. You cannot accidentally produce a spurious GWAS association through confounding in the same way that confounding can produce spurious correlations in expression data from diseased tissue.
The practical evidence supports this reasoning. Analyses of approved drugs consistently show that targets with genetic evidence for human disease relevance have roughly twice the success rate through clinical development compared to targets without it. That is a large, robustly replicated effect, and it justifies placing genetic evidence at the top of the evidence hierarchy.
The Important Caveat
Genetic evidence is the strongest prior, but it is not a sufficient condition for target validity, and absence of genetic evidence is not evidence of absence. GWAS studies require large well-powered cohorts, and those cohorts are not evenly distributed across diseases, populations, or disease subtypes. Rare diseases often lack the patient numbers for genome-wide statistical power. Diseases with different prevalence patterns across ancestries may be systematically underrepresented in predominantly European-ancestry biobanks that have historically dominated GWAS infrastructure. A target that lacks genetic evidence in the current literature might reflect the limits of what has been studied, not the limits of what is biologically relevant.
There is also the locus-to-gene ambiguity problem. Many GWAS hits implicate genomic regions in high linkage disequilibrium containing multiple genes. The actual functional gene is often unclear without additional fine-mapping and experimental work. Genetic evidence can point at a genomic region while leaving substantial ambiguity about which protein in that region is the intended therapeutic target. Acting on a GWAS hit without resolving this ambiguity means you may be pursuing the wrong protein, even when the genetic signal itself is valid.
A Hierarchy, Not a Filter
The practical implication is that genetic evidence should be treated as the highest-weight evidence tier in target scoring, not as a binary filter. A target that lacks current genetic support but has strong functional and literature evidence is not automatically excluded; it carries higher uncertainty that should be reflected in the confidence interval of its composite score, not in a hard exclusion from consideration.
Treating genetic evidence as a filter rather than a weight eliminates opportunities in precisely the disease areas where genetic studies are underpowered, which often correlates with high unmet medical need and lower competitive density. A rare disease indication with no large GWAS -- because the patient population is too small to power one -- is not a domain where target selection should be impossible, only one where the evidence base must draw more heavily on functional and mechanistic data while being explicit about the resulting uncertainty.
Functional Evidence as Complement, Not Substitute
Functional evidence -- knockout phenotypes, CRISPR screen results, pathway enrichment analysis, protein interaction network position -- occupies the second tier in most prioritization frameworks, and appropriately so. Functional evidence is experimentally generated and therefore can be designed to address specific questions about target biology. A well-designed CRISPR screen in a disease-relevant cell model can provide direct functional validation that no amount of GWAS analysis can replace.
The limitation is causal ambiguity. Functional evidence in a cell model or animal model does not guarantee that the same biology operates in the human patient population. Target validity failures in clinical development often involve a disconnect between what was observed in preclinical models and what happens in the human disease context. Functional evidence reduces uncertainty but does not eliminate it in the way that human genetic evidence does.
The strongest prioritization profiles combine both: a target with a human genetic association that has been functionally characterized in relevant experimental models and has consistent literature support provides converging evidence from three independent approaches. Each evidence type is confirmatory of the others, and their convergence substantially reduces the probability that any one of them reflects an artifact.
Literature Evidence: Breadth and Bias
Literature evidence -- the depth and recency of published work on a target's connection to a disease -- occupies the third tier. It is real evidence: consistent independent replication of a target-disease association across multiple research groups and experimental approaches is meaningful signal. But it carries the systematic biases of the published literature: selection for positive results, concentration of research effort on previously studied targets, and citation patterns that amplify high-visibility findings.
The right use of literature evidence is as a complement to genetic and functional evidence, not as a primary ranking criterion. A target with extensive literature support but no genetic or functional evidence may be well-supported within the literature's reference frame while remaining unvalidated in the ways that most predict clinical success. Weighting literature evidence appropriately requires maintaining awareness of this distinction.
Dynamic Evidence, Dynamic Rankings
One implication of treating evidence types as weighted contributions rather than a fixed hierarchy is that rankings should change as the evidence base changes. A target that today lacks genetic support but has strong functional data might acquire GWAS evidence when a larger cohort study is published. When that happens, the target's ranking should update automatically -- not because someone remembered to check, but because the scoring system is designed to incorporate new evidence as it arrives.
This is the difference between a target list that was compiled and a target ranking that is maintained. Compiled lists become stale; maintained rankings reflect the current state of knowledge. For programs that run on twelve- to eighteen-month timelines, a target assessment that cannot be easily updated to incorporate new published evidence is outdated before the program ends. Building the analytical workflow with update cycles in mind is not an optional refinement; it is a practical requirement for using evidence-based prioritization effectively over the duration of a program.