By Brittany Trang, Ph.D.
In the high-stakes world of biotechnology, the "postmortem"—the analysis of a failed clinical trial—is often treated like a closely guarded corporate secret. Companies typically prefer to bury the results of unsuccessful drug development, hoping to pivot toward the next promising candidate without dwelling on the ghosts of trials past. However, in an industry increasingly driven by machine learning and predictive modeling, Verge Genomics is challenging this cultural norm.
Verge, a company founded over a decade ago with the ambitious goal of mapping the genetic networks behind neurodegenerative diseases, recently faced a significant hurdle: the failure of its Phase 1b clinical trial for an ALS drug. Rather than sweeping the data under the rug, Verge is choosing transparency, repositioning its failed trial as a foundational benchmarking dataset for the future of AI-driven drug discovery.
The Core Narrative: Turning "Trash" into "Treasure"
The adage that "one man’s trash is another man’s treasure" has found a modern, clinical application at Verge Labs. For Alice Zhang, co-founder and CEO of Verge Genomics, the failure of a drug trial is not merely a loss of capital and time; it is a dense, high-fidelity source of biological data.
Verge was built on the premise that neurodegenerative diseases like Parkinson’s, ALS, and Alzheimer’s are not caused by single genetic mutations, but by complex, interconnected networks of genes. By mapping these networks, Verge aimed to identify drug targets that traditional, linear discovery methods might miss. While the company has seen successes—including the nomination of two targets to Eli Lilly’s internal pipeline earlier in 2024—the recent setback in its own clinical pipeline serves as a stark reminder of the limitations of current predictive modeling.
By choosing to publish a detailed postmortem of the ALS trial, Zhang is attempting to shift the industry standard. "While the temptation is strong, when a trial doesn’t meet the anticipated end points, to kind of look away and not talk about it, we think there are a lot of learnings that can come — not just for us, but for the field and for ALS broadly — that’s really important to share," Zhang told STAT. "That’s not done very often."
Chronology of a Clinical Setback
To understand the magnitude of this pivot, one must look at the timeline of Verge Genomics’ journey:
- 2013-2015: Verge Genomics is founded with a focus on using patient-derived data and AI to build "all-in-human" models of neurodegenerative disease, bypassing the reliance on traditional animal models.
- 2021-2022: The company makes significant strides in target identification, building a proprietary database of gene expression profiles from patient tissues.
- 2024: Verge announces that Eli Lilly has officially nominated two novel targets identified by the Verge platform for its internal drug development pipeline, signaling institutional confidence in the platform’s target discovery capabilities.
- Mid-2024: The company advances its lead candidate into a Phase 1b trial for ALS.
- Late 2024: The Phase 1b trial fails to meet primary endpoints. The trial is marked by significant attrition, with approximately one-third of participants dropping out due to an inability to tolerate the drug.
- Post-Trial (Current): Verge publishes a comprehensive technical breakdown of the trial failure, focusing on patient tolerance and the specific biological signatures that differed between the predicted model and the human clinical reality.
Supporting Data: Why the AI "Mismatch" Matters
The failure of the ALS trial provides a rare, transparent look into the "black box" of AI drug discovery. In the postmortem published on the Verge Labs blog, the company outlines several key findings regarding the gap between their preclinical AI models and the clinical trial environment.
1. The Tolerance Threshold
One of the most significant hurdles identified was the high dropout rate. The drug, while perhaps effective at the cellular level in silico, proved problematic in vivo. This indicates a discrepancy between the simulated physiological response and the actual systemic metabolic reactions observed in patients.
2. Genetic Network Complexity
Verge’s AI is designed to account for network dynamics rather than single-gene interactions. However, the trial revealed that while the AI correctly identified "hubs" of disease activity, it failed to predict how those hubs would interact with systemic side effects in a complex human organism.
3. Benchmarking as an Asset
Verge is now using the data from this trial to "re-train" its models. By feeding the exact biological signatures of patients who dropped out back into the system, the company is refining its predictive algorithms to better filter for "tolerability" as a primary variable in future candidate selection.
Official Responses: The Philosophy of Transparency
Alice Zhang’s decision to go public with the failure is an outlier in the biotech sector. In an interview with STAT, she emphasized that the culture of silence around failed trials is detrimental to scientific progress.
"If we want to build an industry where AI actually works," Zhang explained, "we have to be willing to show where it fails. We are treating this dataset as a benchmark for the next generation of our models. If other companies did this, we would have a much more robust landscape for ALS research. We would stop repeating the same mistakes."
Industry observers have noted that this level of transparency could actually increase investor trust. By demonstrating an ability to learn from failure, Verge is positioning itself not just as a drug manufacturer, but as a data-first technology firm. In the eyes of partners like Eli Lilly, the value of the Verge platform may now be higher because it has been "stress-tested" against real-world clinical failure, resulting in a more refined algorithm.
Implications for the AI-Biotech Industry
The implications of Verge Genomics’ decision are far-reaching. If the industry moves toward a model of "open-science" postmortems, it could fundamentally alter the economics of drug discovery.
1. Standardization of Failure
Currently, there is no standard for how AI companies report trial failures. If regulators like the FDA or organizations like the ALS Association encourage or mandate the publication of "negative data," it would create a massive, anonymized repository of what doesn’t work, significantly reducing the "discovery tax" that companies pay by hitting the same dead ends.
2. AI Model Generalizability
The Verge case study highlights that AI is only as good as the clinical feedback loop. The industry is currently moving from a phase of "hype" to a phase of "integration." Integration requires acknowledging that human biology is not a static data set. The future of AI in medicine lies in models that can adapt to human clinical failure, rather than just optimizing for preclinical success.
3. Patient Advocacy
For patients with ALS, a terminal diagnosis with limited treatment options, the news of a failed trial is devastating. However, there is a silver lining in transparency. By sharing exactly why the trial failed, Verge provides patients and advocacy groups with a more accurate understanding of the biological hurdles remaining in the field, fostering a sense of partnership rather than mystery.
Conclusion: A New Era for Predictive Medicine
The journey of Verge Genomics from clinical setback to data-driven recovery is a testament to the maturation of the AI-in-medicine field. By transforming a failed ALS trial into a benchmarking dataset, Verge is doing more than just salvaging a project; they are providing a blueprint for how biotech companies should handle the inevitable failures of drug development.
As the industry continues to grapple with the limitations of predictive modeling, the willingness to share, analyze, and learn from these failures will distinguish the leaders of the next decade. If AI is to truly revolutionize health and medicine, it must be allowed to fail—and more importantly, it must be allowed to learn from those failures in the light of day.
Brittany Trang, Ph.D., covers AI in health and medicine. For more insights on the intersection of technology and patient care, subscribe to the weekly AI Prognosis newsletter. Follow Brittany on Threads, Mastodon, and Bluesky.
