Skip to content
ContentLora

    Tip: press / anywhere to search.

    Analysis

    Do AI-designed drugs work? What the clinical evidence shows

    The clinical evidence for AI-designed drugs rests on few trials, and the nearest computationally designed candidate awaits an FDA decision in early 2027.[1][2] Early data suggest AI-derived molecules pass phase I safety tests unusually often but do no better than average in phase II.[3][4]

    Editor reviewedStrict sourcingUpdated AI for scienceHealth and medicineLife sciencesArtificial intelligence
    Show:

    AI for drug discovery has attracted billions of dollars, including a $2.1 billion round for isomorphic-labs in May 2026.[5] The question that matters to patients is narrower: do molecules designed with AI work better, or reach patients faster? This page separates what trials show from what is hoped. It covers medical evidence and is not medical advice.

    What the trials show

    Few novel AI-discovered or AI-designed drugs have reached human trials, according to the authors of the rentosertib trial described below.[1]

    A 2024 analysis of AI-native biotech pipelines in Drug Discovery Today found that AI-discovered molecules succeeded in phase I 80–90% of the time, above historic industry averages.[3] In phase II the success rate was about 40%, comparable to historic averages, on a limited sample.[4]

    The rentosertib trial

    Rentosertib is described by its developers as an AI-generated inhibitor of TNIK, a target in idiopathic pulmonary fibrosis (IPF) that was also identified with generative AI.[6] Its phase 2a trial, published in Nature Medicine in June 2025, randomised 71 patients to three doses or placebo for 12 weeks.[7]

    The primary endpoint was safety: the share of patients with at least one treatment-emergent adverse event was similar across arms.[8] On a secondary endpoint, forced vital capacity rose by a mean 98.4 ml at the highest dose (95% CI 10.9 to 185.9) against a mean change of −20.3 ml on placebo (95% CI −116.1 to 75.6).[9] The authors concluded the drug warrants larger and longer trials.[10]

    On 7 July 2026 Insilico Medicine announced the start of that larger trial: a randomised, double-blind, placebo-controlled phase III study expected to enrol 320 IPF patients in China and run for 52 weeks.[11] It is registered as CTR20262475 and NCT07687459.[12]

    The rentosertib result is a signal, not proof of efficacy. The trial was small and short, the lung-function benefit was a secondary endpoint, and the confidence intervals are wide. It does show that an AI-nominated target and molecule can reach a controlled patient trial without obvious safety problems. The phase III trial, a year long and more than four times larger, is the first real test of whether the lung-function signal holds.

    The zasocitinib case

    On 14 September 2026 Takeda said the FDA had accepted its application for zasocitinib, an oral TYK2 inhibitor for plaque psoriasis, under priority review, with a decision due in the first quarter of 2027.[2] The filing rests on phase 3 data from nearly 3,000 patients.[13] The molecule came from a Nimbus Therapeutics programme that worked with Schrödinger from 2016 and used machine learning plus physics-based free energy perturbation to assess more than 13,000 compounds computationally.[14]

    If zasocitinib is approved, some will call it the first AI-designed drug. That label is debatable: the key method was physics-based simulation, with machine learning in a supporting role. Counts of “AI drugs” vary widely depending on whether they include target discovery, molecule generation, or computational screening.

    Where structure AI fits

    AlphaFold 3 reported far greater accuracy for protein–ligand interactions than standard docking tools, the kind of prediction drug designers rely on.[15] Isomorphic Labs, led by an AlphaFold Nobel laureate, has partnerships with Novartis, Eli Lilly and Johnson & Johnson.[16][17][18] On 29 September 2026 its president wrote that the company’s preclinical data was helping it gear up for clinical development, without announcing a trial.[19]

    Better binding predictions mainly speed up the early, preclinical stages. Most drugs fail later because they do not work well enough in people or cause side effects, which structure prediction alone does not solve. That is consistent with the pattern of strong phase I but average phase II results.

    What would settle it

    Likely signals over the next 12–18 months: the FDA decision on zasocitinib (due Q1 2027), the first clinical trial starts from Isomorphic Labs, and progress in rentosertib’s phase III trial, which began in July 2026. An updated pipeline analysis with more phase II and phase III readouts would show whether the phase II rate rises above industry norms. Our view, with moderate confidence: AI will be shown to shorten discovery timelines before it is shown to raise the odds that a drug works.

    Competing views

    AI is already improving the odds

    AI-derived molecules have shown 80–90% phase I success, well above historic averages, and a placebo-controlled phase 2a trial of an AI-generated drug reported a lung-function signal. Better molecules should translate into more approvals.[3][9]

    No proof yet where it counts

    Phase II success for AI-derived molecules is about 40%, in line with industry norms, and the furthest efficacy data come from a 12-week, 71-patient trial whose efficacy measures were secondary endpoints. Efficacy, not drug-likeness, is the bottleneck.[4][7][9]

    Computation helps, but labels mislead

    A leading candidate for approval, zasocitinib, came from physics-based simulation combined with machine learning, not generative AI alone. Counting 'AI drugs' depends on definitions.[14][2]

    Questions readers ask

    Has an AI-designed drug been approved?

    The closest computationally designed candidate, zasocitinib, has an FDA decision due in the first quarter of 2027; the furthest published generative-AI drug, rentosertib, completed a phase 2a trial and began a phase III trial in July 2026.[2][7][11]

    How do AI-discovered drugs perform in trials?

    A 2024 analysis found 80–90% success in phase I and about 40% in phase II, the latter similar to industry averages and based on a small sample.[3][4]

    What did the rentosertib trial show?

    Adverse-event rates were similar across arms, and forced vital capacity rose 98.4 ml at the top dose versus a 20.3 ml fall on placebo over 12 weeks; the authors call for larger, longer trials.[8][9][10]

    Sources

    Each numbered claim is a statement we checked against the sources listed with it. Status shows how well established it is.

    1. [1]

      The rentosertib paper notes that few novel AI-discovered or AI-designed drugs have reached human clinical trials. confirmedas of 2025-06-03

    2. [2]

      On 14 September 2026 Takeda said the FDA had accepted its new drug application for zasocitinib (TAK-279), an oral TYK2 inhibitor for moderate-to-severe plaque psoriasis, under priority review with a target action date in the first quarter of 2027. confirmedas of 2026-09-14

    3. [3]

      A 2024 analysis of AI-native biotech pipelines in Drug Discovery Today found an 80–90% phase I success rate for AI-discovered molecules, higher than historic industry averages. confirmedas of 2024-06-01

    4. [4]

      The same 2024 analysis found a phase II success rate of about 40% for AI-discovered molecules, on a limited sample, comparable to historic industry averages. confirmedas of 2024-06-01

    5. [5]

      On 12 May 2026 Isomorphic Labs announced a $2.1 billion funding round led by Thrive Capital, with Alphabet and GV among existing investors and MGX, Temasek, CapitalG and the UK Sovereign AI Fund as new investors. confirmedas of 2026-05-12

    6. [6]

      Rentosertib is described by its developers as a first-in-class small-molecule TNIK inhibitor generated with AI, against a target that was also identified using generative AI. confirmedas of 2025-06-03

    7. [7]

      A phase 2a trial of rentosertib, published in Nature Medicine in June 2025, randomized 71 patients with idiopathic pulmonary fibrosis to one of three doses or placebo for 12 weeks. confirmedas of 2025-06-03

    8. [8]

      The rentosertib phase 2a trial's primary endpoint, the share of patients with at least one treatment-emergent adverse event, was similar across all arms (70.6% to 83.3%). confirmedas of 2025-06-03

    9. [9]

      In the rentosertib phase 2a trial, forced vital capacity rose by a mean 98.4 ml (95% CI 10.9 to 185.9) at the highest dose versus a mean change of −20.3 ml (95% CI −116.1 to 75.6) with placebo, a secondary endpoint. confirmedas of 2025-06-03

    10. [10]

      The rentosertib trial authors concluded the drug warrants investigation in larger and longer clinical trials. confirmedas of 2025-06-03

    11. [11]

      On 7 July 2026 Insilico Medicine announced the start of a phase III trial of rentosertib, a randomised, double-blind, placebo-controlled study expected to enrol 320 patients with idiopathic pulmonary fibrosis in China over 52 weeks. confirmedas of 2026-07-07

    12. [12]

      Insilico lists the rentosertib phase III trial under registry numbers CTR20262475 and NCT07687459. confirmedas of 2026-07-07

    13. [13]

      Takeda said the zasocitinib application rests on pivotal phase 3 data from nearly 3,000 patients. confirmedas of 2026-09-14

    14. [14]

      Zasocitinib came from a Nimbus Therapeutics TYK2 programme that, with Schrödinger from 2016, used machine learning and physics-based free energy perturbation to assess more than 13,000 compounds computationally. confirmedas of 2026-09-25

    15. [15]

      The AlphaFold 3 paper reported far greater accuracy for protein–ligand interactions than state-of-the-art docking tools and higher antibody–antigen accuracy than AlphaFold-Multimer. confirmedas of 2024-05-08

    16. [16]

      Isomorphic Labs has strategic partnerships with Novartis, Eli Lilly and Johnson & Johnson. confirmedas of 2026-05-12

    17. [17]

      Demis Hassabis is chief executive of Isomorphic Labs and Max Jaderberg is its president, according to the company's May 2026 funding announcement. confirmedas of 2026-05-12

    18. [18]

      The 2024 Nobel Prize in Chemistry went half to David Baker for computational protein design and half jointly to Demis Hassabis and John Jumper for protein structure prediction. confirmedas of 2024-10-09

    19. [19]

      On 29 September 2026 Isomorphic Labs' president wrote that the company's preclinical data was helping it "gear up for clinical developments"; the post did not announce a clinical trial. confirmedas of 2026-09-29

    Revision history (2)
    1. Page created.
    2. Added rentosertib's phase III start (July 2026) and Isomorphic Labs' clinical status as of September 2026.

    Created Oct 10, 2026. Last reviewed by an editor on Oct 10, 2026. Next scheduled review: Jan 10, 2027.

    Cite this page

    "Do AI-designed drugs work? What the clinical evidence shows." ContentLora, updated Oct 10, 2026. https://contentlora.com/analysis/ai-designed-drugs-evidence

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.

    Keep exploring

    ☰ Contents