Acentric Research is produced with AI tools under human review. The authors are not financial professionals, and nothing here is investment advice or a recommendation to buy or sell any security. AI analysis can be wrong; check every claim against its cited source.
This Brief summarizes the Junction Report AI × Drug Discovery: The Intersection Ecosystem – November 2026, published 2026-11-17. It adds no new research; facts and figures are the report’s, as of 2026-11-17, and the sources the report cites are footnoted here. AI × Drug Discovery was the reference Junction of the Landscape Report and was not ranked there.
Overview
Drug discovery is the early part of drug development: choosing a biological target, then finding and refining a molecule that acts on it. Artificial intelligence (AI) is used here in four ways: models trained on data that rank molecules or targets; algorithms that propose new molecules; programs that predict a protein’s three-dimensional shape from its sequence, such as AlphaFold; [1] and large pretrained models adapted to tasks such as these. [2]
What stands out is the gap between claims and results. FDA approved lirafugratinib on 2026-09-23. Its developer names machine learning as one ingredient of its platform, and no filing the report found says AI designed the molecule; no approved molecule comes from a company that ties its molecule or platform to AI outright. The gain that companies claim is speed in reaching a drug candidate. The report found no sign yet that later, more costly stages go better: a sizeable share of candidates tied to AI have been stopped, and Phase 1 progress, the only later stage compared, could not be told apart from the industry’s. [2] Sources date from 2016 to 2026.
Current state
An approval. FDA approved lirafugratinib (Lyrfigtu) on 2026-09-23 for adults with previously treated cholangiocarcinoma, a bile-duct cancer, with an FGFR2 gene fusion or other rearrangement. The application was filed by Elevar Therapeutics. [3] Relay Therapeutics, which developed the molecule as RLY-4008, licensed it to Elevar in December 2024. [4] The registry lists a Phase 1/2 trial of RLY-4008. [5] In its 2022 annual report Relay wrote that its Dynamo platform “combines advanced machine learning models and molecular dynamics simulations” with laboratory work (company claim); the report found no Relay filing that says AI or machine learning designed this molecule. [6]
Late-stage trials. Three candidates that their companies link to AI, through the molecule or the discovery platform, are in Phase 3 trials, according to ClinicalTrials.gov: rentosertib from Insilico Medicine, for idiopathic pulmonary fibrosis and run in China; bexobrutideg from Nurix Therapeutics; and GB-0895 from Generate Biomedicines. [7] In 2023 Insilico described another of its candidates, ISM5411, now in a Phase 2 trial, as “Designed by AI from scratch” and “developed by Chemistry42, Insilico’s generative AI engine for molecule development” (company claim). [8]
Speed. Six candidates have a company-stated time from the start of a program to a development candidate, the molecule taken into formal safety testing. The median is 11.5 months (10.5 on each candidate’s earliest statement). The comparison is a literature figure of 4.5 years for discovery research, from a 2010 study restated in 2016. [9] Every stated time is a company claim, from companies that chose to publish it. The source for rentosertib’s 18 months is a paper by company staff. [10][2]
Who is involved. Schrödinger describes its software platform as built on “rigorous, physics-based methods, combined with the rapid data processing and scaling advantages of machine learning” (company claim). [11] Large drugmakers work mostly through partnerships: the report found AI partnerships for at least 16 of the 20 largest drugmakers. Examples include Isomorphic Labs with Eli Lilly in 2024, with $45 million upfront and up to $1.7 billion in milestones (company claim), [12] and Nurix with Roche in 2026, with $700 million upfront and up to $2.3 billion in total payments, according to Nurix’s filing, which does not describe the deal in AI terms. [13][14]
What happened to the candidates. The report followed 32 candidates that company documents of the time tied to AI and that were in trials by the end of 2023. By 2026-11-17, 13 (40.6%) had been stopped under the report’s written rules; other cohort and outcome readings give 20.7% to 40.6%. Three of the 13 stops followed trial results. Two were COVID-19 antibodies whose emergency use authorizations FDA revoked, which the report says did not reflect how they were discovered; six were portfolio decisions. [15][16] Of 20 candidates with a known Phase 1 outcome, 12 moved to Phase 2. That is 60%, against 52% for the industry in 2011–2020 and an expected 10.3 of 20 matched by disease area; it is not distinguishable from the industry’s rate, and the sample is too small to rule out a sizeable difference either way. [15][17]
Opportunities
If discovery gets shorter and trials do not, the clinical and regulatory stages, 10.5 years by BIO’s figures, become most of the path to approval. [17] Registry estimates put the end of data collection in Phase 3 trials of five AI-linked candidates, two from companies that name AI as one ingredient of a platform, between April 2028 and October 2029; such estimates often slip. [7][14] On the report’s arithmetic of filing and review times, an FDA decision on a molecule that companies tie to AI outright or to an AI-based platform would come no earlier than 2029–2030, assuming 6–18 months to prepare a filing. [18][2] Too few candidates have resolved Phase 2 outcomes to say whether AI raises Phase 2 success; readouts over 2027–2029 could. [2]
Challenges
- Regulation. FDA’s draft guidance on AI in drug development says that “the use of AI for the purposes of drug discovery is not in the scope of this guidance”. [19] The European Medicines Agency writes that AI in drug discovery “can be low regulatory impact if non-optimal performance only affects the developer”; if results feed evidence for review, non-clinical development principles apply. [20] In the report’s record, an AI-found molecule enters the same review as any other. Iambic Therapeutics discloses that it “cannot predict ... whether the FDA or other regulators will accept AI-derived data or analyses” it may submit. [21]
- Validation. A 2024 perspective states that AI models “are often developed to excel on retrospective benchmarks unlikely to anticipate their prospective performance”. [22]
- Evidence. The claimed speed gain rests on company statements. A recent industry baseline would test it. [2]
- Capital. Among twelve public filers, stated cash runway had a median of 1.55 years from the filing date; the range was 1.0–2.85 years; four state only a 12-month horizon, and none states a doubt about continuing as a going concern. [23]
- Records lag decisions. Nurix decided in August 2026 to stop zelebrudomide, while ClinicalTrials.gov still listed its trial as recruiting on 2026-11-17, so counts of stopped candidates are a floor. [24][2]
Data sources
- ClinicalTrials.gov, for trial phases, status and estimated dates. It lags company decisions. [25]
- Drugs@FDA (openFDA), for approvals. Search by molecule and code name as well as sponsor, because licensed molecules change sponsor. [26][2]
- SEC EDGAR filings: annual reports, quarterly reports, prospectuses and Form 8-K deal notices, for what companies say about their platforms, deals, runway and risks. [27]
- Company releases and annual reports of non-US listed companies, for deal terms and pipeline status. They are company claims. [2]
- FDA and EMA guidance, for how regulators treat AI. [19][20]
- Europe PMC, for peer-reviewed reviews of validation. [28]
- No good source: the role AI played in finding a given molecule can be checked only against the company’s own statements, not against internal records, and no recent public baseline for discovery time was found. [2]
Changelog
- 2026-11-18: First published.
Sources
- Jumper J. et al., “Highly accurate protein structure prediction with AlphaFold,” Nature, 2021-07-15. doi.org(accessed 2026-11-18)
- Acentric Research, “AI × Drug Discovery: The Intersection Ecosystem – November 2026”, 2026-11-17 (the Junction Report this Brief summarizes; see its sections “Method and limitations” and “Open questions”). acentricresearch.com (accessed 2026-11-18)
- U.S. Food and Drug Administration, approval letter for NDA 220425, Lyrfigtu (lirafugratinib), Elevar Therapeutics, 2026-09-23. accessdata.fda.gov (accessed 2026-11-18)
- Relay Therapeutics, Form 10-K for fiscal 2025, filed 2026-02-26 (company claim). sec.gov (accessed 2026-11-18)
- ClinicalTrials.gov, NCT04526106 (RLY-4008, Phase 1/2). clinicaltrials.gov (accessed 2026-11-18)
- Relay Therapeutics, Form 10-K for fiscal 2022, filed 2023 (company claim). sec.gov (accessed 2026-11-18)
- ClinicalTrials.gov, Phase 3 records: rentosertib, NCT07687459; bexobrutideg, NCT07516093; GB-0895, NCT07276724; zovegalisib, NCT06982521; aleniglipron, NCT07654374 (accessed 2026-11-18)
- Insilico Medicine, release of 2023-12-05 on ISM5411, verbatim copy on Kyodo News PR Wire (company claim). kyodonewsprwire.jp (accessed 2026-11-18)
- Schuhmacher A., Gassmann O., Hinder M., “Changing R&D models in research-based pharmaceutical companies,” Journal of Translational Medicine 14:105, 2016, restating Paul S.M. et al. (2010). doi.org (accessed 2026-11-18)
- Ren F. et al., “A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models,” Nature Biotechnology, 2024-03-08 (authors include Insilico Medicine staff). doi.org (accessed 2026-11-18)
- Schrödinger, Form 10-K for fiscal 2025, filed 2026-02-25 (company claim). sec.gov (accessed 2026-11-18)
- Isomorphic Labs, “Isomorphic Labs announces strategic multi-target research collaboration with Lilly,” 2024-01-07 (company claim). storage.googleapis.com (accessed 2026-11-18)
- Nurix Therapeutics, Form 8-K, Item 1.01, filed 2026-06-08. sec.gov (accessed 2026-11-18)
- Acentric Research, deal table and signposts in the AI × Drug Discovery data appendix (computed by the Junction Report’s scripts, not recomputed for this Brief). acentricresearch.com (accessed 2026-11-18)
- Acentric Research, cohort outcomes and Phase 1 → 2 comparison in the AI × Drug Discovery data appendix (computed by the Junction Report’s scripts, not recomputed for this Brief). acentricresearch.com (accessed 2026-11-18)
- U.S. Food and Drug Administration, “Emergency Use Authorization — Archived Information” (entries for bamlanivimab and etesevimab, and for bebtelovimab). fda.gov (accessed 2026-11-18)
- BIO, Informa Pharma Intelligence and QLS Advisors, Clinical Development Success Rates and Contributing Factors 2011–2020, February 2021. go.bio.org (accessed 2026-11-18)
- U.S. Food and Drug Administration, PDUFA Reauthorization Performance Goals and Procedures, Fiscal Years 2023 through 2027. fda.gov (accessed 2026-11-18)
- U.S. Food and Drug Administration, Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products, draft guidance, January 2025. fda.gov (accessed 2026-11-18)
- European Medicines Agency, Reflection paper on the use of artificial intelligence (AI) in the medicinal product lifecycle, EMA/CHMP/CVMP/83833/2023, adopted 9 September 2024. ema.europa.eu (accessed 2026-11-18)
- Iambic Therapeutics, prospectus (Form 424B4), filed 2026-10-15 (company claim). sec.gov (accessed 2026-11-18)
- “Data-centric challenges with the application and adoption of artificial intelligence for drug discovery,” Expert Opinion on Drug Discovery, 2024 (abstract only). doi.org (accessed 2026-11-18)
- Acentric Research, runway as stated in twelve annual reports, in the AI × Drug Discovery data appendix (computed by the Junction Report’s scripts, not recomputed for this Brief). acentricresearch.com (accessed 2026-11-18)
- Nurix Therapeutics, Form 10-Q for the quarter ended 2026-08-31, filed 2026-10-13. sec.gov (accessed 2026-11-18)
- National Library of Medicine, ClinicalTrials.gov. clinicaltrials.gov (accessed 2026-11-18)
- U.S. Food and Drug Administration, openFDA Drugs@FDA bulk download. download.open.fda.gov (accessed 2026-11-18)
- US Securities and Exchange Commission, “Search Filings” (EDGAR). sec.gov (accessed 2026-11-18)
- Europe PMC, REST API. ebi.ac.uk (accessed 2026-11-18)