Universities as Catalysts
INFORMED PROSTATE CANCER SUPPORT GROUP NEWSLETTER
Universities as Catalysts: How AI Could Accelerate Development of Non-Patentable Medicines
The dandelion root extract trial stalled not because of bad science or lack of funding—it failed because there's no profit motive to organize and run a costly clinical trial for a non-patentable substance. Universities like UCSD, armed with world-class AI infrastructure and mission-driven research cultures, could break this logjam by creating hybrid nonprofit-academic models that use machine learning to dramatically reduce trial costs and timelines. The precedent exists. The capability exists. What's needed is institutional commitment and dedicated funding to prove the model works.
The Dandelion Root Lesson: Why Big Pharma Won't Solve This
The Windsor, Ontario dandelion root story is instructive not for what it failed to accomplish, but for what it reveals about pharmaceutical economics. A promising natural therapy for blood cancers received regulatory approval for human trials, secured charitable foundation funding, attracted international patient interest—and then came to a halt.
The problem wasn't scientific doubt. It wasn't funding for early research. It was the cost and logistics of running a clinical trial: recruiting 30 patients is expensive. Monitoring their safety is time-consuming. Data management, statistical analysis, regulatory compliance—these are the expensive parts, and they don't get cheaper because the underlying therapy is inexpensive to make.
For a pharmaceutical company, these trial costs are justified by future patent protection and market monopoly. For a non-patentable substance, there is no downstream revenue to offset trial expenses. This creates a structural void: promising but unmarketable candidates languish in the gap between preliminary research (which charities will fund) and clinical proof (which only commercial incentives typically support).
This is where universities enter the picture—not as cheerleaders, but as operators.
What Universities Actually Have
When we talk about universities having "AI capabilities," we often mean computational power and talented researchers. But that's only part of what UCSD and comparable research institutions bring to the table:
1. AI Infrastructure at Scale
UCSD is home to the Institute for Learning-enabled Optimization at Scale (TILOS), part of a $220 million NSF National AI Research Institutes investment. This isn't a single lab with a GPU—it's an institutional commitment to multidisciplinary AI research including optimization, data integration, and responsible computing. The institution already has the computational infrastructure to run the kind of machine learning models that would accelerate drug discovery and trial design.
Additionally, UCSD researchers have demonstrated concrete progress in AI drug discovery. The POLYGON system, developed at UCSD's School of Medicine, uses trained neural networks to simulate chemistry and generate candidate drug compounds for cancer that would normally take thousands of experiments to identify. The system can design novel molecules in silico in days—work that would traditionally take months.
2. Deep Drug Development Expertise
UCSD houses the Center for Drug Discovery Innovation (CCDI), a dedicated unit that "helps UC San Diego researchers build and advance drug discovery projects by providing expert advice, facilitating access to research resources, and networking them with other researchers on and beyond the campus." This isn't just a research group—it's infrastructure designed to support the full pipeline from concept to preclinical development.
The Division of Extended Studies offers professional training in drug discovery and development, clinical trial administration, pharmacology, and ADME (Absorption, Distribution, Metabolism, Excretion)—expertise that would be invaluable for designing efficient nonprofit clinical trials.
3. Proximity to Patients and Clinical Sites
UC San Diego is directly affiliated with UC San Diego Moores Cancer Center, one of the nation's leading academic medical centers. The institution has established clinical trial infrastructure, institutional review boards, and most importantly, patient populations seeking treatment. This is a significant logistical advantage that commercial contractors lack—a university-based trial doesn't require negotiating access to patient populations; the medical center is part of the same institution.
4. Mission-Driven Culture and Non-Profit Status
Unlike commercial firms, universities are nonprofits with explicit missions to advance science and serve public health. This means institutional incentives align with completing research even when there's no commercial payoff. A university can absorb marginal costs on a nonprofit trial because the overhead (buildings, IT, HR) is already paid for by research and education funding.
How Universities Could Deploy AI to Reduce Trial Costs
Here's where the economics become favorable for non-patentable drugs. AI can compress clinical trial timelines and reduce recruitment costs—the two most expensive components. A university-based model could make a crucial difference:
AI-Powered Patient Recruitment and Matching
The Windsor dandelion trial recruited 5 patients over 5 years despite needing only 30. A modern AI-powered recruitment system could have transformed this. UCSD could deploy trained machine learning models to:
- Scan Moores Cancer Center's electronic health records to identify eligible patients with blood cancers
- Automatically flag patients whose disease characteristics match trial criteria
- Prioritize enrollment efforts on the most promising candidates
- Use natural language processing to extract clinical information from unstructured medical notes
Research shows that AI-powered recruitment systems achieve 93–96% accuracy and can identify eligible patients in minutes rather than hours of manual review. For a trial needing 30 patients, this could compress enrollment from 5 years to 6–12 months.
AI-Optimized Trial Design
Before enrolling a single patient, AI can simulate the entire trial in silico using historical data. UCSD researchers could:
- Optimize inclusion/exclusion criteria to expand the eligible population without compromising safety
- Predict optimal dose ranges and treatment duration
- Design adaptive protocols that adjust in real-time based on safety and efficacy signals
- Reduce sample size requirements through Bayesian statistical methods and synthetic control arms
This could reduce the trial from Phase I/II (30 patients, 10 months, substantial cost) to a streamlined design requiring fewer patients and shorter duration.
Real-Time Safety Monitoring
UCSD's AI infrastructure could provide continuous monitoring of trial data, flagging safety signals automatically rather than waiting for interim analyses. Digital biomarkers and real-time data entry systems could reduce administrative burden on trial sites.
Data Analysis and Reporting
Natural language processing and machine learning models can rapidly analyze trial results, identify biomarkers predictive of response, and prepare statistical summaries for regulatory submission—work that normally takes statisticians weeks to complete.
The Funding Model: Hybrid Academic-Nonprofit Structure
The missing piece in the dandelion equation wasn't science or capability—it was sustained, dedicated funding. A university-based model could solve this through a hybrid structure that already exists in other contexts:
Who Pays For What
- Charitable foundations: Support early preclinical research and clinical trial design (as they did for dandelion)
- University infrastructure: Provides computational capacity, clinical site access, and administrative overhead (already funded through research grants and tuition)
- Dedicated nonprofit entity: A nonprofit arm of the university (or independent nonprofit partner) sponsors and manages the trial, absorbing trial-specific costs
- Public funding: ARPA-H, NIH, or government health agencies could fund specific trials addressing public health gaps
This model isn't theoretical. Multiple universities and nonprofits already operate this way. Genethon, a French nonprofit biotherapy R&D organization, has sponsored or partnered in four Phase I/II clinical trials and leads over a dozen preclinical projects for rare genetic diseases, all without patent-based revenue streams. Their model relies on foundation funding, government grants, and the willingness of academic partners to contribute infrastructure.
Cost Reduction Through Infrastructure Sharing
A key advantage of the university model is that marginal costs are much lower than for a startup biotech or commercial CRO. The infrastructure expenses—clinical trial coordinators, institutional review board review, regulatory compliance, data management—exist at Moores Cancer Center regardless. A non-patentable trial can leverage this existing capacity.
Research on nonprofit drug development shows that costs can be dramatically lower when leveraging existing infrastructure. For instance, the Council for Scientific and Industrial Research's Open Source Drug Discovery project in India has spent $35 million on tuberculosis drug discovery and clinical trial preparation—amounts that would be consumed in overhead alone by commercial approaches.
Precedent: What Other Universities and Nonprofits Have Done
University-led, nonprofit-funded drug development is not speculative. Multiple institutions demonstrate that it works:
Founded by MIT-trained entrepreneur Jaykumar Menon, the Open Source Pharma Foundation operates under a "design backward from scale" model, working with existing generic medicines and repurposing them for new diseases through systematic clinical trials. The foundation has progressed multiple candidates to Phase IIB/III trials for conditions like tuberculosis and infectious disease. Funding comes from foundations, government agencies, and philanthropists—not patents. The key insight: by starting with existing off-patent drugs and leveraging university and hospital infrastructure in disease-endemic regions, they can conduct rigorous clinical trials at a fraction of commercial cost.
Since 1990, Genethon has sponsored or partnered on multiple Phase I/II clinical trials for rare genetic diseases, launching therapies that later became commercial products through partnerships with biotech companies. Genethon operates as a nonprofit research organization funded by the Telethon charity in France. Its success demonstrates that universities and nonprofits can design and execute sophisticated early-stage clinical trials without patent protection, and that commercial partners will later invest in trials that nonprofits validate.
India's government-funded CSIR OSDD has built a tuberculosis drug discovery pipeline with 54 molecules in development, including candidates in clinical trials, using a purely open-source model where volunteers contribute computational and experimental work worldwide. The project receives government funding ($35 million allocated) and conducts clinical trials in publicly-funded hospitals in partnership with hospital staff and contracted pharmaceutical specialists. This demonstrates that large-scale, rigorous clinical development is possible without traditional pharma economics.
Nonprofits like CHDI Foundation (Huntington's disease) and the Michael J. Fox Foundation partner with Contract Research Organizations like Evotec, where the CRO covers costs in exchange for co-ownership of successful IP. This model shows that nonprofit sponsors can work with specialized research firms on a cost-sharing basis. Universities could adopt similar arrangements, contracting specialized trial management to CROs while providing the clinical site infrastructure and patient population.
What all these examples share: When universities or nonprofits sponsor trials, they can absorb costs that would be prohibitive for a for-profit biotech because they operate on mission and public trust rather than return-on-equity.
What UCSD Specifically Could Do (A Concrete Proposal)
UCSD has the pieces. Here's how they could be assembled into a systematic capability:
1. Establish a "Nonprofit Clinical Translation Initiative"
Create a dedicated unit within UCSD's Center for Drug Discovery Innovation (or as an affiliated nonprofit) focused specifically on advancing non-patentable candidates into clinical trials. Staff this with trial design experts, regulatory specialists, and AI engineers.
2. Deploy POLYGON and Related AI Tools for Trial Optimization
Use UCSD's existing POLYGON platform and other trained machine learning models to:
- Design optimized Phase I trial protocols before enrollment
- Identify best patient populations through retrospective data analysis
- Simulate trial outcomes and predict success probability
3. Build a Charity Trial Consortium
Partner with charitable foundations (like those that funded dandelion research) to create a shared pool of funding for trial conduct. Instead of each foundation independently funding a trial, they collectively fund an institutional capacity that runs multiple trials sequentially—achieving economies of scale.
4. Leverage Moores Cancer Center Clinical Infrastructure
Use UCSD's medical center as the trial site, reducing site activation and management costs. The medical center's institutional review board, research coordinators, and patient populations are already in place.
5. Recruit Patient Participants Using AI-Powered Matching
Deploy AI systems to identify eligible patients from Moores Cancer Center's records and flag them for recruitment—the approach that would have transformed the dandelion trial from 5 enrolled over 5 years to rapid enrollment.
6. Share Results Openly
Commit to publishing all data, trial designs, and results in open-access venues. This creates two benefits: (1) foundations and donors see their money creating public knowledge, not proprietary secrets, and (2) the knowledge becomes available to other institutions and nonprofits running similar trials globally.
Estimated Impact
If UCSD established this infrastructure and took on a trial like dandelion root extract, AI-optimized design and recruitment could:
- Reduce enrollment from 5 years to 12–18 months
- Reduce trial cost from commercial estimates ($5–10M for Phase I/II) to $1–2M through infrastructure leverage
- Enable completion with 20–25 patients instead of 30, reducing exposure and timelines further
- Provide regulatory-quality evidence suitable for FDA submission
Barriers and How to Address Them
Barrier 1: Institutional Incentives
Problem: Universities are incentivized by research grants and publications, not by running clinical trials. Clinical trial work is demanding and does not generate the prestige of fundamental research.
Solution: Reframe clinical trial leadership as a core research mission. Provide faculty appointment credit and promotion consideration for clinicians and scientists who lead nonprofit trials. Highlight the public health impact in promotion and awards materials.
Barrier 2: Regulatory and Legal Complexity
Problem: Running Phase I/II trials requires extensive regulatory knowledge and legal infrastructure that universities don't typically maintain in-house.
Solution: Hire or contract specialized regulatory and legal expertise. This is a known cost and can be amortized across multiple trials. Alternatively, partner with CROs that specialize in nonprofit trials and can provide regulatory support on a cost-sharing basis.
Barrier 3: Funding Reliability
Problem: Foundation funding for individual trials is unpredictable. A multi-year trial needs funding commitment beyond what most foundations typically provide.
Solution: Establish a dedicated fund through major donors or endowment, with annual grants sufficient to run 1–2 trials concurrently. This provides stability and allows universities to recruit and retain trial staff. ARPA-H and NIH could fund this as a demonstration project.
Barrier 4: Patient Recruitment in Rare Diseases
Problem: Many non-patentable candidates address rare diseases where patient populations are small and dispersed.
Solution: Use AI to aggregate and identify patients across multiple sites. Partner with disease-specific patient organizations (like IPCSG for prostate cancer) to conduct recruitment. Use digital/remote trial designs that reduce travel burden.
Why This Matters Beyond Dandelion Root
The dandelion extract case is a template, not an exception. There are potentially dozens of promising natural products, repurposed generic drugs, and academic discoveries that languish because they can't be patented and thus can't be commercialized through traditional pharma channels:
- Lemongrass extract (also studied by Windsor researchers for cancer)
- Existing drugs with new disease indications that have limited commercial upside
- Botanical and traditional medicine compounds with preclinical evidence
- Therapies for ultra-rare diseases serving populations too small for commercial development
- Drugs addressing diseases predominantly affecting low-income populations with limited profit potential
If universities like UCSD created institutional capacity for nonprofit clinical trials, they could address an entire category of unmet medical needs that the commercial system structurally ignores. This is not about competing with big pharma—it's about occupying a space that big pharma will never serve because the economics don't work.
For prostate cancer specifically, this could accelerate work on natural therapies, alternative dosing regimens of existing drugs, and precision medicine approaches that lack commercial incentive.
Next Steps
If UCSD leadership wanted to move forward, a reasonable sequence would be:
- Year 1: Conduct a feasibility study identifying 3–5 promising non-patentable candidates. Estimate trial costs and timelines using AI-optimized design. Identify foundation funding partners.
- Year 2: Launch pilot trial on one candidate. Demonstrate that AI-driven recruitment and design reduce timelines and costs.
- Years 3+: Scale to 1–2 trials annually. Build reputation as the academic home for nonprofit drug development. Attract dedicated funding.
Funding for Year 1 might require a $500K–$1M grant from a foundation committed to health innovation. The payoff would be proof that universities can bridge the gap between academic discovery and clinical proof—and that AI makes this gap-bridging economically feasible.
https://ucsd.edu/research-innovation/artificial-intelligence.html
https://cddi.ucsd.edu/
https://today.ucsd.edu/story/simulated-chemistry-new-ai-platform-designs-tomorrows-cancer-drugs
https://extendedstudies.ucsd.edu/courses-certificates/sciences/therapeutic-discovery-and-development
https://intuitionlabs.ai/articles/ms-ai-drug-development-programs
https://pmc.ncbi.nlm.nih.gov/articles/PMC3743608/
https://www.mckinsey.com/alumni/news-and-events/global-news/alumni-news/2024-02-are-open-source-principles-the-key-to-saving-lives
https://www.sciencedirect.com/science/article/pii/S135964462500011X
https://cen.acs.org/pharmaceuticals/drug-discovery/Open-source-drug-discovery-takes-aim-at-malaria-and-neglected-diseases/97/i5
https://pmc.ncbi.nlm.nih.gov/articles/PMC6249613/
https://journals.plos.org/plosntds/article?id=10.1371/journal.pntd.0001827
https://cen.acs.org/business/outsourcing/Contract-research-organizations-profit-nonprofitsdrug/97/i31
https://pmc.ncbi.nlm.nih.gov/articles/PMC5369032/
https://pmc.ncbi.nlm.nih.gov/articles/PMC9155596/
https://www.sciencedirect.com/science/article/abs/pii/S1472979211001028
https://ardigen.com/ai-in-biotech-lessons-from-2025-and-the-trends-shaping-drug-discovery-in-2026/
https://www.sciencedirect.com/science/article/abs/pii/S0031699725075118
Comments
Post a Comment