Open Source Drug Discovery 2.0

A hybrid IP and incentive framework for democratizing translational AI drug research

Abstract

Recent advances in artificial intelligence and data integration promise to compress discovery timelines, yet drug development remains hamstrung by soaring costs and attrition. Eroom’s law — the observation that the inflation‑adjusted cost of developing a new drug roughly doubles every nine years — has persisted despite high‑throughput screening, combinatorial chemistry, and computational design[1], and the average R&D cycle still spans 10–15 years and over US$2.6 billion per medicine, with only ∼12 % of candidates reaching approval[2]. This Perspective proposes Open Source Drug Discovery 2.0 (OSDD-2), a socio-technical and economic framework that harnesses open data, AI-driven tooling, and a hybrid intellectual property (IP) model to democratize early discovery while preserving investible exclusivity for late-stage development. OSDD‑2 builds on lessons from pre‑AI open‑source initiatives such as India’s Council for Scientific and Industrial Research Open Source Drug Discovery (CSIR‑OSDD)[3], the Open Source Malaria consortium[4] and the Open Source Pharma Foundation[5], but introduces a governed “IP‑gating” mechanism that transitions an openly developed asset into a field‑limited exclusive licence once a predefined scientific milestone is achieved. We outline the conceptual architecture, compare OSDD-2 with both pre-AI open approaches and the traditional closed model, articulate desiderata and guiding principles, and illustrate the approach through a case study targeting phosphoglycerate dehydrogenase (PHGDH) for Alzheimer’s disease [6]. Finally, we discuss how this hybrid model could catalyse a shift from siloed heroics to transparent relays in drug discovery, drawing inspiration from the cystic fibrosis (CFTR) success story.

Introduction

The past two decades have seen astonishing progress in computing and biology. High-throughput sequencing, cloud computing, and AI models, such as AlphaFold, have dramatically accelerated hypothesis generation and data interpretation. Yet, the productivity of pharmaceutical R&D continues to decline: Eroom’s law notes that the inflation‑adjusted cost to bring a new therapy to market doubles roughly every nine years[1]. Large pharmaceutical companies and venture‑backed biotech start‑ups invest billions to secure a single composition‑of‑matter patent, but nine of ten projects still fail in clinical trials[2]. The conventional “closed” model rewards secrecy and winner‑take‑all outcomes; R&D teams operate behind non‑disclosure agreements, negative data are rarely shared, and duplication abounds. Conversely, a first wave of open‑source drug discovery movements — including CSIR‑OSDD (2008) for tuberculosis[7], Open Source Malaria (2011)[4], and the Open Source Pharma Foundation (2015)[5] — demonstrated that diverse contributors can collaborate on neglected disease targets without patents. These early initiatives made high‑value data public and attracted thousands of participants, but few yielded clinically viable molecules. Lack of sustainable funding, unclear IP positions, and absence of a pathway to regulatory sponsorship limited their impact[8].

In parallel, the success of cystic fibrosis (CF) modulators demonstrates the power of multi-stakeholder collaboration in a closed setting. By partnering with a small biotechnology company, the Cystic Fibrosis Foundation (CFF) pioneered a venture philanthropy model in the late 1990s, directly funding high-throughput screening for CFTR modulators. When Vertex Pharmaceuticals ran parallel trials on multiple corrector–potentiator combinations, an elexacaftor–tezacaftor–ivacaftor “triple” regimen went from synthesis to FDA approval in about three years. Trikafta improved lung function by ~14% and enabled ~90% of patients to achieve near-normal respiratory capacity. The median life expectancy for children on modulators may approach 82 years. The CFF subsequently reinvested US$3.3 billion in royalties into further research and development. The CFTR story underscores that sustained funding, iterative experimentation, and seamless hand‑offs across academia, foundations, biotech, and pharma can yield cures.

Conceptual framework: the three open sources and IP‑gating

OSDD‑2 seeks to reconcile openness with investability through three interlocking “open sources”:

1. Open Evidence: disease maps, omics data, assays, protocols, and negative results are released under FAIR (findable, accessible, interoperable, and reusable) conditions. This concept aligns with CFDE’s ambition to unify heterogeneous datasets and ensures that contributors build upon each other’s work. Transparent evidence reduces redundancy and increases reproducibility.
2. Open Tooling: analytic workflows, AI models, and simulation environments are published with code and documentation. These tools range from gene–disease network algorithms to generative chemistry models and agentic AI for experiment planning. Publicly accessible tooling democratizes participation and facilitates verification.
3. Open IP‑Gating: the core composition‑of‑matter hypotheses and initial lead structures are disclosed on a ledger under a non‑exclusive research licence. All contributions to the open phase are time‑stamped and attributable. When a project reaches a pre-specified milestone (e.g., in vivo proof-of-concept), a gate is triggered: a not-for-profit foundation or DAO selects a team to form a spin-out company. This spin-out receives a field-limited, exclusive licence to advance the asset through IND-enabling studies and clinical development. In return, the foundation retains equity and royalty rights, ensuring that value flows back to the contributor network. This IP‑gating mechanism preserves investibility while honouring open contributions.

The OSDD-2 platform thus functions like a relay: open evidence and tools accelerate early discovery, the gate hands off a de-risked asset to a private entity for late-stage development, and returns capital to seed new open projects. A patient‑centric foundation or DAO oversees governance, milestone definitions, and dispute resolution.

Comparison of models

The table below contrasts three paradigms: pre‑AI open‑source drug discovery (e.g., CSIR‑OSDD[3] and Open Source Malaria[4]), the traditional closed model and the proposed OSDD‑2. OSDD‑2 aims to combine the inclusiveness of open projects with the capital leverage of private spin‑outs.

Desiderata for OSDD‑2

For OSDD‑2 to succeed, the following criteria must be satisfied:

1. Composable, attributable IP — Every data point, code component, and chemical scaffold must be citable with an immutable provenance trail. The licence stack defines how open contributions transition to exclusivity and codifies obligations for forks and derivatives.

2. Cash‑backed, milestone incentives — Contributors should be rewarded for verifiable outcomes, not just effort. Bounties (e.g., target-engagement assays, PK measurements) provide modest cash payments and micro-royalties, funded by philanthropy or early-stage investors.

3. Academic credit that counts — Contributions to open projects should yield citable DOIs, protocol references, and authorship on resulting publications. This aligns with tenure metrics and encourages academic participation.

4. Clear bridge to IND — The platform must partner with or create a public‑benefit CRO capable of acting as an IND sponsor. The DAO does not file NDAs; the spin‑out carries regulatory and liability responsibilities.

5. Robust governance — A foundation or DAO with patient‑advocate representation oversees rules, disputes, and treasury management, preventing capture by any single interest.

6. FAIR data and reproducibility — All data and methods adhere to FAIR principles; negative results and failure modes are recorded. Standardized metadata models, such as CFDE’s C2M2, enable cross-project queries.

7. User‑centric design — The platform must offer intuitive interfaces for data submission, AI analysis, bounty posting, and progress visualisation. Participation should be driven by utility, not just ideology.

The ten principles of OSDD‑2

1. Radical transparency: core hypotheses, data, and negative results are posted to a public ledger, lowering the barrier for others to contribute and reproduce findings.

2. Open ≠ free: value is created through openness, but must be captured. Bounty payments, micro‑royalties, and equity ensure sustainability.

3. Hybrid IP by design: open evidence and tooling coexist with gated exclusivity. This resolves the tension between sharing and investability.

4. Milestone‑driven incentives: rewards correspond to completed assays or validated hypotheses, not to time spent.

5. Patient‑centric governance: patient communities hold formal seats in prioritising targets and evaluating spin‑out proposals.

6. Pragmatic partnership with industry: pharma can monitor open programmes and compete to license or acquire the de‑risked asset after the gate. This complements rather than replaces closed R&D.

7. Interoperable standards: all data and tools adhere to community standards (e.g., C2M2 metadata). Comparable assays enable meaningful head‑to‑head comparisons.

8. Democratised access to AI tools: the platform provides access to cutting‑edge models and workflows so that contributors without large computational resources can participate.

9. Forks with obligations: contributors may fork a programme but must honour original attribution and licensing terms.

10. Relays beat heroes: collaborative hand‑offs from academic discovery to commercial development shorten timelines and diffuse risk, as demonstrated by the CFTR relay.

Case study: PHGDH as the first OSDD‑2 programme

Alzheimer’s disease (AD) remains an area of high unmet need. A 2025 study in Cell discovered that the metabolic enzyme phosphoglycerate dehydrogenase (PHGDH) plays a non‑enzymatic transcriptional role in late‑onset AD. The authors demonstrated that PHGDH overexpression in astrocytes promotes amyloid pathology, whereas knockdown suppresses it [6]. Mechanistically, PHGDH enhances transcription of IKKa and HMGB1, leading to impaired autophagy. A blood‑brain‑barrier‑permeable small‑molecule inhibitor targeting PHGDH’s transcriptional function reduced amyloid pathology and improved behavioural deficits in mice and human brain organoids[6]. PHGDH’s RNA and protein levels correlate with cognitive decline and increase before clinical diagnosis[9], suggesting both therapeutic and diagnostic potential.

Our group has designed novel PHGDH inhibitors and filed a provisional patent. Under OSDD-2, we propose posting the core scaffold structures, docking models, and initial activity data as open evidence. Contributors can claim bounties for tasks such as reproducing binding affinity using surface plasmon resonance, assessing blood‑brain barrier permeability, performing ADME/toxicity assays, and evaluating efficacy in organoids or mouse models. Each milestone carries a modest payment (e.g., US$500–1,000) and a micro-royalty of 0.01–0.05%. Upon demonstrating in vivo efficacy and acceptable safety, the programme would trigger the gate. A spin-out, e.g., OSDD-Studio-PHGDH, would secure an exclusive license for AD indications, raise venture capital, and serve as an IND sponsor. Contributors would share equity proportional to their ledgered contributions and be co‑authors on the initial publications.

Discussion: from CFTR to AI‑enabled open source

The CFTR journey illustrates how long-term collaboration and creative funding can overcome a once-fatal disease. The CFF’s decision to fund high‑throughput screening and partner with biotech overcame pharmaceutical reluctance. Vertex’s bold strategy of running multiple combination trials in parallel accelerated the timeline from molecule to approval, leading to the triple therapy Trikafta. Nearly 90 % of CF patients now have an effective treatment, and life expectancy is approaching normal. This success did not emerge from a single lab but from a relay across academia, philanthropy, industry, and regulators. OSDD-2 seeks to replicate this relay model across diseases, utilizing open evidence and AI to increase the number of shots on goal and IP-gating to ensure there is still a pot of gold to justify late-stage investment.

Pre‑AI open‑source drug projects provided valuable precedents but lacked sustainable funding and clear paths to market. CSIR-OSDD mobilized thousands of participants to identify tuberculosis targets, but its public-domain outputs were unable to attract venture backing. [3] The Open Source Malaria consortium has demonstrated open notebook science and rapid iteration, but has yet to deliver a clinical candidate [4]. The Open Source Pharma Foundation advocates for end-to-end open pharma but still operates mainly as a thought leader. [5] OSDD‑2 retains openness while recognising that exclusivity at the right juncture unlocks capital.

The integration of AI and data commons further differentiates OSDD‑2. Platforms like the NIH Common Fund Data Ecosystem (CFDE) harmonise multi‑program datasets and provide search and analysis tools. AI-driven generative chemistry, protein modeling, and agentic workflows can reduce the time from target identification to lead optimization. However, without transparent data sharing and reproducibility, AI models risk learning from flawed or biased datasets. OSDD-2’s open evidence layer ensures that AI outputs can be audited and collectively improved.

Conclusion

Drug discovery stands at a crossroads. Traditional closed models are delivering diminishing returns despite record-high R&D spending [1][2], while early open-source experiments lacked investibility. By combining open data and AI tools with hybrid IP-gating and milestone-based incentives, OSDD-2 offers a pragmatic path forward. It invites participation from academic labs, computational scientists, patient groups, foundations, CROs, and biotechs. Through bounties and transparent credit, it rewards verifiable contributions. Through a governed gate and spin‑out mechanism, it preserves the ability to raise the hundreds of millions needed for clinical trials. As the PHGDH case study illustrates, OSDD‑2 can transform promising hypotheses into IND‑ready assets while documenting every step. The model is not a panacea; regulators, investors, and funders will need to adapt. But if we are to bend Eroom’s law and realise the full promise of AI in medicine, we must shift from isolated heroics to open, accountable relays. The first Open Source Drug Discovery Conference, scheduled for Birmingham in 2026, will provide an opportunity to test these ideas in practice. We invite the community to join the relay and help turn open science into approved therapies.

References

1. Scannell JW, Blanckley A, Boldon H & Warrington B. Diagnosing the decline in pharmaceutical R&D efficiency. Nat. Rev. Drug Discov. 11, 191–200 (2012).

2. PhRMA. Research and Development policy Framework, https://www.phrma.org/policy-issues/research-development (2024).

3. Council of Scientific and Industrial Research Team India Consortium. Open Source Drug Discovery (OSDD) initiative. https://en.wikipedia.org/wiki/Open_Source_Drug_Discovery (accessed 2025).

4. Open Source Malaria Consortium. Battling disease with open. https://sparcopen.org/impact-story/open-source-malaria-consortium/ (accessed 2025).

5. Open Source Pharma Foundation (OSPF). https://www.ospfound.org/about.html (accessed 2025).

6. Chen JY. Cystic Fibrosis: a Blueprint for Rare Disease Cures. PODCAST https://open.spotify.com/episode/59rJdE2vQEkSzwnLfrLR6X?si=90bf2c874a494108 (2025).

7. Chen J, Hadi F, Wen X, Zhao W, Xu M, Xue S, Lin P, Calandrelli R, Richard JLC, Song Z, Li J, Amani A, Liu Y, Chen X, Zhong S. Transcriptional regulation by PHGDH drives amyloid pathology in Alzheimer’s disease. Cell 188, 3513–3529.e26 (2025).

8. The NIH Common Fund Data Ecosystem (CFDE) overview. http://cfdeconnect.org (2025).