Genetic Evidence and Phase III Drug Outcomes
Diagnosing Drug Failure Modes from Cross-Design Evidence Across Ten Disease Domains
Mendelian randomization (MR) evidence is associated with drug success, but when a drug fails despite genetic support, or succeeds without it, MR alone cannot say why. Observational and MR effect sizes for 41 mechanism families, each a drug target or exposure paired with a disease indication across ten disease domains, were standardized to Cohen’s d with a fixed threshold of 0.10, and a deterministic rule classified each family from the pattern of the two. Under the preregistered analysis the rule classifies 24 of 32 scored families correctly (75.0%; outcome-permutation p = 0.005). The observational evidence adds no information beyond MR status; the eight misclassified families split into five with null MR evidence and an approved drug and three with supportive MR evidence and a failed drug.