Loading
What changed between revision 5 and revision 8, compared over the article's markdown source with its frontmatter — so a change to status, evidence or reversibility shows up as the edit it is.
Older · revision 5
Import
Re-import of content/ai-drug-discovery.md — the file changed after the initial import. Not an edit.
23.4 kB
Revision 5
Revision 8
lines 8–14 → 8–147 unchanged lines not shown
categories: ["foundations", "longevity"]categories: ["foundations", "longevity"]tags: ["machine learning", "drug development", "generative chemistry", "target discovery", "clinical trials", "pharmacology"]tags: ["machine learning", "drug development", "generative chemistry", "target discovery", "clinical trials", "pharmacology"]summary: "The use of machine learning to select drug targets, generate candidate molecules and predict failure, so far demonstrably faster at making compounds than at making medicines."summary: "The use of machine learning to select drug targets, generate candidate molecules and predict failure, so far demonstrably faster at making compounds than at making medicines."updated: "2026-10-28"updated: "2026-10-30"humanEvidence: "AI-derived small molecules have reached phase 1 and 2 trials, with a first phase 3 begun in 2026; sponsors of an mRNA therapy built on algorithm-selected neoantigens report a phase 3 in melanoma met its endpoints, and no drug from either route is approved."humanEvidence: "AI-derived small molecules have reached phase 1 and 2 trials, with a first phase 3 begun in 2026; sponsors of an mRNA therapy built on algorithm-selected neoantigens report a phase 3 in melanoma met its endpoints, and no drug from either route is approved."access: "Platforms are proprietary to pharmaceutical companies or licensed to them, and some models and datasets are public; no molecule designed this way is available outside a trial, though AI-suggested uses of approved drugs are."access: "Platforms are proprietary to pharmaceutical companies or licensed to them, and some models and datasets are public; no molecule designed this way is available outside a trial, though AI-suggested uses of approved drugs are."reversibility: "context"reversibility: "context"lines 50–57 → 50–5735 unchanged lines not shown
language models reading that literature directly. The most consequential public case islanguage models reading that literature directly. The most consequential public case isBenevolentAI's identification of baricitinib, an approved rheumatoid arthritis drug,BenevolentAI's identification of baricitinib, an approved rheumatoid arthritis drug,as a plausible COVID-19 treatment in early 2020.[^richardson2020] Baricitinib went on to beas a plausible COVID-19 treatment in early 2020.[^richardson2020] Baricitinib went on to beauthorized for that use. It is a genuine result and a narrow one: the system reranked knownauthorized for that use. The result was real but narrow: the system reranked known pharmacologypharmacology rather than producing a molecule.and did not produce a molecule. **Molecule generation** trains models over SMILES strings, molecular graphs or three-dimensional**Molecule generation** trains models over SMILES strings, molecular graphs or three-dimensionalposes, then steers them with reinforcement learning or conditions them on a binding pocket.poses, then steers them with reinforcement learning or conditions them on a binding pocket.lines 59–66 → 59–661 unchanged line not shown
claiming designed inhibitors in weeks rather than months.[^zhavoronkov2019]claiming designed inhibitors in weeks rather than months.[^zhavoronkov2019] **Property prediction** estimates absorption, metabolism, cardiac ion-channel liability and toxicity**Property prediction** estimates absorption, metabolism, cardiac ion-channel liability and toxicitybefore synthesis. This is the least glamorous application and probably the most useful, because itbefore synthesis. It is the least glamorous of these tasks and probably the most useful, becausekills bad series cheaply.it kills bad compound series cheaply. **Screening triage** trains a classifier on a modest set of measured compounds and applies it to**Screening triage** trains a classifier on a modest set of measured compounds and applies it tolibraries too large to test. A graph neural network trained on a few thousand molecules testedlibraries too large to test. A graph neural network trained on a few thousand molecules testedlines 81–91 → 81–9414 unchanged lines not shown
coordinating work across contract laboratories. These are where developers of general-purposecoordinating work across contract laboratories. These are where developers of general-purposelanguage and agent models have entered the field, by a different route from Alphabet's, which spunlanguage and agent models have entered the field, by a different route from Alphabet's, which spunits drug-design work out into a separate company. Anthropic reported in October 2026 on researchits drug-design work out into a separate company. Anthropic reported in October 2026 on researchpartnerships with six drug developers — BigHat Biosciences, Bristol Myers Squibb, Divergence Labs,partnerships with six drug developers (BigHat Biosciences, Bristol Myers Squibb, Divergence Labs,Enveda, Generate Biomedicines and The Column Group — and on programmes of its own, having acquiredEnveda, Generate Biomedicines and The Column Group) and on programmes of its own, having acquiredMithril Medicine, a one-person company working on two rare forms of muscular dystrophy whose founderMithril Medicine, which it describes as a one-person company working on two rare forms of muscularhad used the company's Claude models to coordinate experiments across eleven externaldystrophy whose founder had used the company's Claude models to coordinate experiments across elevenlaboratories.[^anthropic2026] Anthropic's stated reasoning is that the bottlenecks are distributedexternal laboratories.[^anthropic2026] Mithril's technical report dates the acquisition to July 2026and names its author, Cody Tranbarger, as the founder; it also lists a team of paid consultants andadvisers, and every experiment in it was performed by contract researchorganizations.[^tranbarger2026] Anthropic's stated reasoning is that the bottlenecks are distributedthrough the whole development process rather than concentrated in molecule design.through the whole development process rather than concentrated in molecule design. ## Development history## Development historylines 98–104 → 101–1076 unchanged lines not shown
{ "year": "2020", "title": "First AI-designed molecule in the clinic", "text": "DSP-1181, from Exscientia and Sumitomo Dainippon Pharma, enters a phase 1 trial in Japan for obsessive-compulsive disorder. It was later discontinued." }, { "year": "2020", "title": "First AI-designed molecule in the clinic", "text": "DSP-1181, from Exscientia and Sumitomo Dainippon Pharma, enters a phase 1 trial in Japan for obsessive-compulsive disorder. It was later discontinued." }, { "year": "2020", "title": "Antibiotic found by screening model", "text": "The Collins laboratory identifies halicin using a graph neural network, the first widely cited antibiotic hit attributed to deep learning." }, { "year": "2020", "title": "Antibiotic found by screening model", "text": "The Collins laboratory identifies halicin using a graph neural network, the first widely cited antibiotic hit attributed to deep learning." }, { "year": "2021", "title": "AlphaFold2 released", "text": "Predicted structures for most of the human proteome are made public, changing what structure-based design can be attempted." }, { "year": "2021", "title": "AlphaFold2 released", "text": "Predicted structures for most of the human proteome are made public, changing what structure-based design can be attempted." }, { "year": "2021–2022", "title": "AI-derived target and molecule reach humans", "text": "Insilico's ISM001-055, a TNIK inhibitor for idiopathic pulmonary fibrosis, enters first-in-human testing after both target and compound were nominated computationally." }, { "year": "2021-2022", "title": "AI-derived target and molecule reach humans", "text": "Insilico's ISM001-055, a TNIK inhibitor for idiopathic pulmonary fibrosis, enters first-in-human testing after both target and compound were nominated computationally." }, { "year": "2024", "title": "Consolidation", "text": "Exscientia and Recursion merge, combining two of the largest AI-first drug discovery companies after neither had produced an approved medicine." }, { "year": "2024", "title": "Consolidation", "text": "Exscientia and Recursion merge, combining two of the largest AI-first drug discovery companies after neither had produced an approved medicine." }, { "year": "2026", "title": "First AI-designed small molecule enters phase 3", "text": "Insilico Medicine begins a 320-patient randomized placebo-controlled phase 3 trial of rentosertib in idiopathic pulmonary fibrosis in China, the furthest any generatively designed small molecule has gone." }, { "year": "2026", "title": "First AI-designed small molecule enters phase 3", "text": "Insilico Medicine begins a 320-patient randomized placebo-controlled phase 3 trial of rentosertib in idiopathic pulmonary fibrosis in China, the furthest any generatively designed small molecule has gone." }, { "year": "2026", "title": "Neoantigen therapy meets phase 3 endpoints", "text": "Merck and Moderna report that intismeran autogene, an individualized mRNA therapy encoding up to 34 algorithm-selected neoantigens, met its recurrence-free survival endpoint with pembrolizumab in resected melanoma." } { "year": "2026", "title": "Neoantigen therapy meets phase 3 endpoints", "text": "Merck and Moderna report that intismeran autogene, an individualized mRNA therapy encoding up to 34 algorithm-selected neoantigens, met its recurrence-free survival endpoint with pembrolizumab in resected melanoma." }lines 131–137 → 134–14026 unchanged lines not shown
Merck and Moderna announced that intismeran autogene, an individualized neoantigen mRNA therapyMerck and Moderna announced that intismeran autogene, an individualized neoantigen mRNA therapycovered under [[personalized-mrna-cancer-vaccines]], met its primary endpoint of recurrence-freecovered under [[personalized-mrna-cancer-vaccines]], met its primary endpoint of recurrence-freesurvival and a key secondary endpoint of distant metastasis-free survival when added tosurvival and a key secondary endpoint of distant metastasis-free survival when added topembrolizumab in a 1,137-patient trial in resected stage IIB–IV melanoma. The claim is topline, frompembrolizumab in a 1,137-patient trial in resected stage IIB-IV melanoma. The claim is topline, froman interim analysis, and detailed data had not been released as of late August 2026.[^merck2026]an interim analysis, and detailed data had not been released as of late August 2026.[^merck2026]Each course is built by sequencing the patient's tumour and encoding up to 34 of its mutations,Each course is built by sequencing the patient's tumour and encoding up to 34 of its mutations,selected by what Moderna describes as a proprietary algorithm developed with Merck.[^moderna2023]selected by what Moderna describes as a proprietary algorithm developed with Merck.[^moderna2023]lines 140–158 → 143–1772 unchanged lines not shown
phase 3 depends entirely on where the label's boundary is drawn.phase 3 depends entirely on where the label's boundary is drawn. The newest claims for the approach are not near patients at all. Anthropic's October 2026 reportThe newest claims for the approach are not near patients at all. Anthropic's October 2026 reportmixes molecule-design results — roughly 35,000 antibody designs generated for Divergence Labs inmixes molecule-design results with operational ones. On the design side it lists roughly 35,000about a month, about 1,010 antibody variants made and tested with BigHat Biosciences over about tenantibody designs generated for Divergence Labs in about a month, about 1,010 antibody variants madeweeks, a shortened genetic component of a CAR T-cell therapy that Generate Biomedicines aims toand tested with BigHat Biosciences over about ten weeks, and a shortened genetic component of a CARextend to [[cancer|solid tumours]] — with operational ones, among them a manufacturing-documentT-cell therapy that Generate Biomedicines aims to extend to [[cancer|solid tumours]]. Thereview at Enveda that the company says cut human review time by 89%, and an assessment ofoperational work includes a manufacturing-document review at Enveda that the company says cut humanclinical-trial plans at Bristol Myers Squibb. Every figure is reported by the companies themselves,review time by 89%, and an assessment of clinical-trial plans at Bristol Myers Squibb. Every figureand Anthropic describes the work as early and mostly preclinical: nothing from it has entered ais reported by the companies themselves, and Anthropic describes the work as early and mostlyhuman trial, and the earliest first-in-human date the company offers, late 2028, is conditional onpreclinical: for the most part its drugs have not reached human testing, though the Bristol Myersits candidates continuing to hold up. Its furthest-advanced internal programme, a bispecificSquibb work concerns an early-stage drug already in clinical trials. The earliest first-in-human date Anthropic offersantibody for a form of congenital muscular dystrophy, rests on a pilot study in mice whosefor its own programmes, late 2028, is conditional on its candidates continuing to hold up. Itsaccompanying technical report states that it was not peer reviewed and that the experiments werefurthest-advanced internal programme, a bispecific antibody for a form of congenital muscularneither randomized nor blinded.dystrophy, rests on a pilot study in mice whose accompanying technical report states that it was notpeer reviewed and that the experiments were neither randomized nor blinded. That report sets out what the programme has and has not shown. In LAMA2-related congenital musculardystrophy the protein that ties the sheath around each muscle fibre to a receptor in the fibre'smembrane is missing; a related laminin takes its place but holds on weakly, and the antibody ismeant to clamp the two together, one arm on the substitute laminin and one on beta-dystroglycan. Twoformats were studied in depth, a bivalent mouse version and a tetravalent humanized one. Both boundtheir targets at nanomolar affinity and held dystroglycan-overexpressing cells on a laminin fragmentin culture. In dyW mice dosed from birth for six weeks, the tetravalent group developed about twicethe evoked forelimb force and grip strength of vehicle-treated mice, while the bivalent one had nodetectable effect on force. The limitations the report lists are substantial: groups of four toseven mice enrolled in sequence, no prespecified endpoint, and a force difference that gaveP = 0.057 with litters rather than individual mice as the unit of analysis, the smallest value sevenlitters allow. Survival was not an endpoint and histology was not assessed. Claude, the authors write, searched the literature, pulled the contractlaboratories' reports into tables, re-analysed and plotted the data and drafted the text under theirdirection, and made errors that they catalogue.[^tranbarger2026] > [!caution] What "AI-discovered" means> [!caution] What "AI-discovered" means> No regulator recognizes the category, and companies apply the label to everything from a molecule> No regulator recognizes the category, and companies apply the label to everything from a molecule> drawn by a generative model to a conventional campaign in which a classifier filtered one plate.> drawn by a generative model to a conventional campaign in which a classifier filtered one plate.lines 212–220 → 231–23953 unchanged lines not shown
measured in decades. Compressing discovery from four years to one changes little when the limitingmeasured in decades. Compressing discovery from four years to one changes little when the limitingstep is a trial nobody knows how to design. Better preclinical models, whether [[organoids]] orstep is a trial nobody knows how to design. Better preclinical models, whether [[organoids]] orsimulation approaches such as [[human-digital-twins]], plausibly matter more tosimulation approaches such as [[human-digital-twins]], plausibly matter more to[[healthspan|healthspan research]] than faster chemistry does — though the simulation side is[[healthspan|healthspan research]] than faster chemistry does. The simulation side, however, isleast developed in exactly the systems aging research needs, metabolism and immunity, and noleast developed in exactly the systems aging research needs, metabolism and immunity, and no modelmodel of a whole person exists.of a whole person exists. ## Risks## Risks lines 270–275 → 289–29549 unchanged lines not shown
[^zhavoronkov2019]: `paper` Zhavoronkov, A. et al. "Deep learning enables rapid identification of potent DDR1 kinase inhibitors." *Nature Biotechnology*, 2019. {DDR1 was a well-precedented target and the compounds resembled known inhibitors; chemists disputed how much of the speed the model explained.}[^zhavoronkov2019]: `paper` Zhavoronkov, A. et al. "Deep learning enables rapid identification of potent DDR1 kinase inhibitors." *Nature Biotechnology*, 2019. {DDR1 was a well-precedented target and the compounds resembled known inhibitors; chemists disputed how much of the speed the model explained.}[^stokes2020]: `paper` Stokes, J.M. et al. "A Deep Learning Approach to Antibiotic Discovery." *Cell*, 2020. {Halicin's activity was confirmed in mouse infection models; it has not entered human trials.}[^stokes2020]: `paper` Stokes, J.M. et al. "A Deep Learning Approach to Antibiotic Discovery." *Cell*, 2020. {Halicin's activity was confirmed in mouse infection models; it has not entered human trials.}[^anthropic2026]: `statement` Anthropic. "Early progress on accelerating drug development with Claude." Company blog post, October 27, 2026, with the accompanying technical report by Tranbarger, C. and Mithril Medicine on a bispecific bridging antibody for LAMA2-related congenital muscular dystrophy, October 26, 2026. {Every result is the companies' own; the technical report is marked not peer reviewed and its mouse pilot was neither randomized nor blinded.}[^anthropic2026]: `statement` Anthropic. "Early progress on accelerating drug development with Claude." Company blog post, October 27, 2026, with the accompanying technical report by Tranbarger, C. and Mithril Medicine on a bispecific bridging antibody for LAMA2-related congenital muscular dystrophy, October 26, 2026. {Every result is the companies' own; the technical report is marked not peer reviewed and its mouse pilot was neither randomized nor blinded.}[^tranbarger2026]: `statement` Tranbarger, C. and Mithril Medicine. "Preclinical Characterization of a Bispecific Bridging Antibody for the Treatment of LAMA2-Related Congenital Muscular Dystrophy." Anthropic technical report, October 26, 2026. {Not peer reviewed; its competing-interests note states that Anthropic acquired Mithril in July 2026, owns the programme and employs the author, who founded the company.}[^xu2025]: `paper` Xu, Z. et al. "A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial." *Nature Medicine*, 2025. {Seventy-one patients across three dose arms and placebo, treated for twelve weeks; a trial that size and length cannot establish clinical benefit in fibrosis.}[^xu2025]: `paper` Xu, Z. et al. "A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial." *Nature Medicine*, 2025. {Seventy-one patients across three dose arms and placebo, treated for twelve weeks; a trial that size and length cannot establish clinical benefit in fibrosis.}[^insilico2026]: `statement` Insilico Medicine. "Insilico Initiates Phase III Clinical Trial for Rentosertib, Its AI-Empowered TNIK Inhibitor for Idiopathic Pulmonary Fibrosis." Press release, July 2026. {Announces trial initiation only, not results; a 320-patient, 52-week randomized placebo-controlled study at Chinese centres.}[^insilico2026]: `statement` Insilico Medicine. "Insilico Initiates Phase III Clinical Trial for Rentosertib, Its AI-Empowered TNIK Inhibitor for Idiopathic Pulmonary Fibrosis." Press release, July 2026. {Announces trial initiation only, not results; a 320-patient, 52-week randomized placebo-controlled study at Chinese centres.}[^merck2026]: `statement` Merck and Moderna. "Merck and Moderna Announce Phase 3 INTerpath-001 Trial of Intismeran Autogene Plus KEYTRUDA Met Endpoints of RFS and DMFS in Patients With Completely Resected Stage IIB–IV Melanoma." Joint press release, August 19, 2026. {A topline interim-analysis announcement; hazard ratios and detailed results were not disclosed with it.}[^merck2026]: `statement` Merck and Moderna. "Merck and Moderna Announce Phase 3 INTerpath-001 Trial of Intismeran Autogene Plus KEYTRUDA Met Endpoints of RFS and DMFS in Patients With Completely Resected Stage IIB–IV Melanoma." Joint press release, August 19, 2026. {A topline interim-analysis announcement; hazard ratios and detailed results were not disclosed with it.}removed, struck through added, underlinedLine numbers count the serialised markdown of each revision, frontmatter included.
256 lines are the same in both revisions, and 201 of them are not shown. Neither revision here is necessarily the current article — read AI drug discovery for the text this wiki stands behind today.