In the rapidly evolving landscape of modern pharmacology, the narrative has been dominated by the meteoric rise of artificial intelligence. From AlphaFold’s protein-folding breakthroughs to generative models promising to slash the time-to-market for new therapeutics, the industry has tethered its future to the power of big data. Yet, in the Bay Area, a company founded in 2002—long before the current generative AI boom—is staking its reputation on a contrarian premise: that the most transformative drug candidates are not waiting to be found in existing training datasets, but are waiting to be constructed from first principles.
Verseon, a pioneer in computational drug discovery, argues that the industry’s reliance on "AI-driven" design often leads to iterative improvements on existing chemical scaffolds rather than true innovation. By utilizing a "Deep Quantum Modeling" platform, the company is attempting to push drug design beyond the limits of known chemical space, aiming to develop therapeutics for major human diseases that have remained elusive to conventional discovery methods.
The Chronology of a Scientific Counter-Movement
Verseon’s journey began in 2002, a time when "computational drug design" was still a nascent field dominated by rule-based systems and rudimentary docking software. While industry peers like Recursion and Exscientia later rode the wave of the AI revolution, Verseon spent two decades refining a physics-based engine.
The company’s philosophy is rooted in a sobering realization about the history of medicinal chemistry. Over the past 150 years, humanity has synthesized roughly $10^8$ drug-like compounds. While massive digital libraries like ZINC-22 now boast tens of billions of "make-on-demand" molecules, Verseon’s leadership argues that these numbers are misleading. Many of these molecules are merely incremental variations—swapping a fluorine atom for a chlorine atom on a familiar scaffold—leading to a "collapse" of genuine chemical diversity.
"Think of the full set of possibilities as a chemical universe," says Adityo Prakash, co-founder and CEO of Verseon. "What humanity has explored is not a planet. It is a few grains of sand."
To expand beyond these grains of sand, Verseon began building a platform that ignores existing libraries during the initial design phase. Instead, it starts at the atomic level: identifying a protein pocket and calculating, through quantum physics, how to arrange a novel structure that binds perfectly within that target.
Supporting Data: The Limitations of Interpolation
The fundamental critique leveled by Verseon against mainstream AI drug discovery is a distinction between interpolation and extrapolation.
The Interpolation Trap
Prakash and his team argue that deep learning models—the backbone of current AI drug discovery—are inherently biased toward the data they have been fed. Because these models are trained on historical medicinal chemistry data, they excel at interpolating—creating variations of what has already been proven to work. This explains why many AI-designed candidates share structural similarities with known inhibitors. A 2022 analysis by the CAS (Chemical Abstracts Service) highlighted this, finding that many clinical candidates from AI-first firms bore striking structural resemblances to established drugs, such as the antipsychotic haloperidol.
The Extrapolation Challenge
To create truly novel drugs, a platform must extrapolate—venturing into regions of "chemical space" where no data exists. This is where Verseon claims its Deep Quantum Modeling platform holds the advantage. By using physics to model electron interactions and binding energies, the system can predict the viability of structures that have never been synthesized by humans or captured in any database.
Official Perspectives: Verseon’s Strategic Pipeline
Verseon’s approach is not just theoretical; it is currently being stress-tested in several high-stakes clinical and preclinical programs.
Precision Oral Anticoagulants (PROAC)
One of the company’s most mature programs is its Precision Oral Anticoagulant (PROAC) series. The goal is to solve a decades-old clinical problem: creating a blood thinner that prevents strokes without the dangerous bleeding risks associated with current standards like warfarin or newer direct oral anticoagulants (DOACs).
Verseon’s lead candidate, VE-1902, is designed as a reversible covalent thrombin inhibitor. By selectively blocking clot formation while leaving the platelet-activating function of thrombin intact, the company hopes to offer a safer profile. As of mid-2026, the company continues to advance this program, following positive data from rodent models and ongoing Phase I investigations.
Targeting Diabetic Eye Disease
Verseon is also taking an "oral-first" approach to diabetic macular edema, a condition typically treated with repeated, invasive anti-VEGF injections into the eye. By targeting plasma kallikrein with its candidate VE-4840, Verseon aims to reduce retinal vascular leakage via an oral pill, a significant shift in the standard of care.
Implications for the Future of Drug Discovery
The debate between AI-driven "tweaking" and physics-based "creation" has profound implications for the future of the pharmaceutical industry. If the industry continues to produce "the same car with a new paint job," it will likely face a diminishing return on investment as "low-hanging" targets are exhausted.
The Role of AI in Verseon’s Workflow
Importantly, Verseon does not reject AI; it relegates it to a secondary, supporting role. Once the physics-based platform has designed a novel, validated core structure, Verseon uses AI to generate variants and optimize the compound’s secondary properties. This, they argue, is the correct application of machine learning: using it to refine and iterate after the primary, creative breakthrough has been achieved.
A New Standard for Novelty
The broader industry is beginning to recognize the need for higher standards. The 2025 review of 71 published cases in drug discovery noted that while ligand-based models often yield derivative structures, structure-based approaches—similar to the ones Verseon pioneered—tend to result in higher novelty. As regulators and investors become more sophisticated, the "novelty" of a chemical scaffold will likely become a key metric for determining the valuation of drug discovery firms.
Conclusion: The Long Game
Verseon’s persistence as a privately held entity for over two decades highlights the capital-intensive and time-consuming nature of true innovation in life sciences. While the rest of the world rushed to catch the generative AI wave, Verseon spent those years building a foundation based on the immutable laws of physics.
As the industry moves toward later-stage clinical trials for AI-designed drugs—such as the recent progress of Insilico Medicine’s rentosertib—the field will soon have the definitive evidence needed to judge these competing methodologies. If Verseon’s physics-first approach can successfully navigate the "valley of death" between preclinical design and market approval, it may prove that the future of medicine lies not in the gargantuan datasets of the present, but in the fundamental quantum calculations of the past.
For now, Verseon stands as a reminder that in the search for the next generation of therapeutics, the most advanced algorithms are no substitute for a deep understanding of the chemical universe. As Prakash succinctly puts it, "The AI is good at doing what it is supposed to do: create variants. But the scientists still have to validate the prediction." In this high-stakes race, the winners will likely be those who can balance the raw predictive power of AI with the structural rigor of molecular physics.
About the Author
