Rethinking Drug Development Failures
Drug development failure is frequently framed as a clinical trial problem. Headlines often cite missed endpoints, overly broad patient populations, or uninformative biomarkers. These explanations highlight where a program failed, but they rarely pinpoint when the failure began. In many cases, the fatal error occurs long before the trial starts, specifically when a team selects a target without sufficient evidence that it drives disease in humans. Even with advanced chemistry and operational efficiency, a program will falter if the chosen target does not cause the disease process.
Technological progress in medicine is undeniable. Tools like protein design, RNA tech, and AI-enabled screening allow companies to build therapeutic candidates at record speeds. These innovations are helpful, but they remain secondary to the core biological question: does this target matter in humans? The industry often treats target selection as a secondary consideration, yet it remains the primary driver of project success or failure. Focusing on the biology before the molecule is the only way to shift the odds of clinical success.
The Flaw of Literature-Based Selection
At 5 Prime, we analyzed over 100,000 clinical trials to identify patterns in failure and success. The data revealed a striking trend. The primary factor determining whether a target reached clinical testing was the volume of existing literature. Teams pursue targets that feel familiar because they appear in published research. Unfortunately, that frequency of publication has no correlation with whether a drug actually succeeds.
This behavior creates a cycle of bias. Relying on the literature is a safe way to build internal support or convince investors, but it does not account for the biological reality of disease. Many published observations are descriptive rather than causal. Researchers often mistake associations for drivers, leading programs to target the downstream consequences of disease rather than the root mechanism. Treating a symptom is not the same as correcting the cause.
Using Genetics to Determine Causation
Human genetics provides a crucial lens for distinguishing between association and causation. Since genetic variants are assigned at conception, they act as a natural form of randomization. This allows scientists to test whether a protein or metabolite truly influences disease outcomes. Mendelian randomization has emerged as a key tool here. It allows researchers to use genetic variants as proxies for biological shifts, essentially performing a natural experiment that avoids the confounding variables of observational biology.
This evidence must be interpreted with precision. Genetic support is not a binary yes or no. A common variant might point to a locus without identifying the specific gene. A rare variant might be informative but lack the statistical reach to prove a broad effect. Well-interpreted genetics helps determine whether a gene is likely causal, whether a drug should inhibit or activate that target, and whether the timing of the disease relates to the biological mechanism. Weak interpretation leads to false confidence, while rigorous analysis prevents the pursuit of fragile hypotheses.
Reframing Progression and Safety
Most genetic studies focus on susceptibility, or who develops a disease. However, many medicines address patients who are already sick. For these patients, progression biology is often more relevant than susceptibility factors. A gene that influences the onset of a disease may not be the same one that drives its worsening over time. Ignoring this distinction can lead to trial failures, especially in autoimmune and neurodegenerative conditions where the factors driving early and late stages often diverge.
Safety is similarly underserved by current early-stage practices. Human genetics can highlight the potential consequences of long-term target modulation by showing what happens when a gene functions differently in the general population. While this is not a substitute for formal toxicology, it does allow teams to identify potential on-target liabilities before the first human is dosed. Incorporating these findings early moves safety planning from a reactive state to a proactive one.
A Higher Standard for Actionability
The goal of integrating causal biology is not just to produce larger evidence packets. The objective is to produce evidence that forces a decision. A strong biological hypothesis should dictate whether a program stops, pauses, or accelerates. It should define the biomarker strategy and refine the patient selection criteria. This is a higher standard than statistical significance alone. It requires evidence that survives the scrutiny of portfolio committees and translates into real clinical benefit. If the industry is willing to commit years of time and millions of dollars to a project, the foundation of that project must be based on a causal human link. Starting with the biology is the only way to ensure the molecules being built have a genuine chance to succeed.