Rentosertib

Rentosertib is an investigational oral small-molecule inhibitor of TNIK (TRAF2- and NCK-interacting kinase) developed by Insilico Medicine for idiopathic pulmonary fibrosis (IPF), a progressive scarring disease of the lungs. Insilico says its PandaOmics engine nominated TNIK as a fibrosis target and its Chemistry42 generative-chemistry engine designed the molecule, and the company describes rentosertib as the first drug candidate in which both the target and the compound came out of generative AI.^[1][2][3] That claim is why the drug is discussed outside pulmonology: it is treated as a test case for whether AI drug discovery shortens the front end of pharmaceutical research.

The molecule carried the development codes ISM001-055 and INS018_055 before the United States Adopted Names Council assigned the generic name rentosertib in March 2025.^[4] It has been through a microdose study in Australia, two Phase 1 studies, a completed Phase 2a trial in China, and, since September 2026, a Phase 3 trial in China.^[1][2][5][6] It is not approved by any regulator. The Phase 3 study is the first trial sized to test clinical benefit, with primary completion estimated for October 2029.^[5]

Rentosertib drew a second round of attention on 7 September 2026, when a paper in Nature Biotechnology reported that six published proteomic aging clocks, applied to serum samples from 42 participants in the completed Phase 2a trial, all predicted a lower biological age in the treated arms than in placebo.^[3] Those were exploratory biomarker analyses in a small cohort, and the authors state plainly that the analysis cannot separate an effect on aging biology from the drug's anti-fibrotic activity.^[3]

Chemistry and target

Rentosertib is a bis-imidazolecarboxamide. Its molecular formula is C27H30FN7O with a molecular weight of about 487.6, its CAS registry number is 2828567-39-9, and PubChem records the name as both a United States Adopted Name and an International Nonproprietary Name.^[7] The medicinal chemistry campaign that produced it was published in the Journal of Medicinal Chemistry in 2024, where the compound appears as compound 4; the series was designed to modulate the conformation of the TNIK gatekeeper residue Met105 and reach the kinase back pocket.^[8] In a kinase panel the compound was most potent against TNIK, with a half-maximal inhibitory concentration of 31 nM.^[1] The 2026 Nature Biotechnology paper gives its half-life as 7 to 11 hours and notes that 60 mg once daily produced two to three times the peak plasma concentration of either 30 mg regimen.^[3]

TNIK sits in the germinal center kinase family, acts as a RAP2 effector, and mediates WNT signaling through direct interaction with the TCF-LEF complex.^[1] TNIK inhibition had been studied in oncology before Insilico's program, notably with the tool compound NCB-0846, but that compound hits additional kinases including CDK2 at 0.1 uM; the 2024 paper argues rentosertib's selectivity and the gap between its anti-fibrotic and cytotoxic concentrations distinguish it.^[1] In cell work, rentosertib suppressed TGF-beta-induced alpha-smooth muscle actin expression in MRC-5 lung fibroblasts with an IC50 of 27 nM, at concentrations far below those affecting cell viability, and showed anti-fibrotic activity in lung, kidney and skin models by oral, inhaled and topical routes.^[1] A separate study using Insilico's robotics laboratory reported that pharmacological TNIK inhibition acts as a senomorphic agent, suppressing the senescent phenotype in IMR-90 cells.^[9]

How the target and the molecule were found

Insilico applied PandaOmics to lung and kidney fibrosis datasets to generate target hypotheses, using what the company calls a time machine approach that trains on data available before a cutoff date. Under the platform's "protein and receptor kinase" setting, combined with filters for novelty, small-molecule druggability and protein class, TNIK scored first among the candidates.^[1]

Chemistry42 then generated leads against the TNIK ATP-binding site. The paper describes 30 generative models running in parallel, sharing structures and scores with each other during generation, followed by a virtual-screening feedback loop and a lead-optimization stage aimed mainly at absorption, distribution, metabolism and excretion properties.^[1] Insilico reports that the work from target discovery to nomination of the preclinical candidate took roughly 18 months, a figure stated in the peer-reviewed paper itself.^[1] The company adds that fewer than 80 small molecules were synthesized and tested along the way, and puts its typical range at 60 to 200 molecules per program; that molecule count is a company figure rather than one reported in the paper.^[10]

The preclinical candidate was nominated in February 2021.^[11]

Clinical development

The Phase 0 study in Australia was the first administration of the compound to humans.^[1] The two Phase 1 studies produced comparable safety and pharmacokinetic profiles in New Zealand and Chinese participants, with the exception that the plasma exposure increase from 60 mg to 120 mg in the Chinese single-dose cohort was slightly steeper than dose proportionality would predict.^[1]

The Phase 3 trial, which Insilico calls GENESIS-IPF-3, is a randomized, double-blind, placebo-controlled study of once-daily rentosertib over 52 weeks. Its primary endpoint is the annual rate of forced vital capacity decline, and its key secondary endpoint is time to first disease-progression event.^[5][6] Insilico announced on 10 September 2026 that the first patient had been dosed at Peking Union Medical College Hospital and that Shanghai Pulmonary Hospital enrolled its first patient the same day, with Zuojun Xu as lead principal investigator and Nanshan Zhong and Chang Chen as co-leading principal investigators.^[6] The trial protocol also collects serum proteomic profiles at weeks 12, 26 and 52 as exploratory measures, carrying the aging-biomarker design of the Phase 2a study into a larger trial.^[5]

Phase 2a results

The Phase 2a trial, published in Nature Medicine on 3 June 2025, randomized 71 adults with IPF at 21 sites in China to placebo (n=17), 30 mg once daily (n=18), 30 mg twice daily (n=18), or 60 mg once daily (n=18) for 12 weeks. Patients already stable on nintedanib or pirfenidone for more than eight weeks were eligible.^[2] The primary endpoint was the share of patients with at least one treatment-emergent adverse event, and rates were similar across arms: 72.2% at 30 mg once daily, 83.3% at 30 mg twice daily, 83.3% at 60 mg once daily and 70.6% on placebo.^[2]

Fifty-five of the 71 patients (77%) completed the placebo-controlled period, with completion lowest in the two arms receiving the highest daily exposure: 12 of 18 (67%) in both the 30 mg twice-daily and 60 mg once-daily groups, against 15 of 17 (88%) on placebo. Sixteen patients discontinued treatment. Seven withdrew because of liver toxicity (0 of 18 at 30 mg once daily, 4 of 18 at 30 mg twice daily, 3 of 18 at 60 mg once daily), and four of those seven were also taking nintedanib. One patient died of heart failure before the week-12 visit, assessed as unrelated to the study drug. Hypokalemia, abnormal hepatic function, diarrhea and raised alanine aminotransferase were among the most frequent adverse events.^[2]

Lung function was a pre-specified secondary endpoint. Over 12 weeks, mean change in forced vital capacity was -20.3 mL on placebo (95% CI -116.1 to 75.6), -27.0 mL at 30 mg once daily, +19.7 mL at 30 mg twice daily, and +98.4 mL at 60 mg once daily (95% CI 10.9 to 185.9). Within the 60 mg group, patients not also taking standard-of-care antifibrotics improved by +187.8 mL (95% CI 68.6 to 306.9), while those taking nintedanib or pirfenidone alongside rentosertib showed no significant improvement.^[2] Insilico's press material also cites a placebo figure of -62.3 mL. That value comes from a company figure caption that excludes one placebo patient and one 30 mg once-daily patient whose screening and baseline FVC readings differed by more than 600 mL, and it is not among the numbers printed in the paper's main text, so it is not used here.^[14]

Lead investigator Zuojun Xu of Peking Union Medical College said rentosertib "represents a truly innovative therapeutic, with both its target identification and molecular design powered by AI".^[14] Zhavoronkov framed the result as data that "warrants further investigation in larger-scale clinical trials of longer duration," which is also the concluding sentence of the paper's abstract.^[2][14]

One number in the record is inconsistent. The Nature Medicine paper says the trial ran at 21 sites in China; the ClinicalTrials.gov record lists 22 locations, and the methods section of the 2026 Nature Biotechnology paper also says 22 sites. The 21-site figure is used above because it is what the trial's own publication reports.^[2][3][12]

The 2026 proteomic aging-clock analysis

The Phase 2a protocol prospectively collected longitudinal serum samples for exploratory biomarker work, profiled on the Olink Explore 3072 platform at baseline and at weeks 2, 4 and 12. Of the 71 randomized patients, 43 consented to proteomic screening and 42 were analyzed after one exclusion for a missing end-of-trial measurement. The resulting cohort was 42 Asian patients with a mean age of 67.1 years, and 2,841 proteins passed quality control.^[3]

Six published proteomic clocks were run on those samples. Four are trained to predict chronological age and two on mortality risk, and they come from five separate publications rather than six:

That composition matters for how the headline is read. Insilico's release describes "six independently-developed proteomic aging clocks," but two of the six are variants of a single model from one paper, and one of the six, ipfP3GPT, is Insilico's own.^[3][20] The other four were built by outside groups, and authors from those groups, Argentieri, Goeminne and Han among them, are coauthors of the 2026 paper.^[3] The release also lists the clock developers as "the leading groups from Harvard, Oxford, PKU, and Insilico," which leaves out the University of Connecticut group behind PAC.^[20]

The four chronological clocks tracked actual age well in this cohort (Spearman's r between 0.70 and 0.84, root mean square error under four years after correction for a constant offset). The two mortality-trained clocks correlated weakly with calendar age (r 0.16 to 0.23) and predicted higher ages, which the authors attribute to the training objective and the disease burden of an IPF cohort.^[3]

Across weeks 2, 4 and 12, three regimens and six clocks, the study made 54 comparisons per arm against placebo. Twenty-one reached statistical significance at a false-discovery-rate threshold of Q < 0.10, and they clustered at week 4, where 11 of 18 comparisons showed a significantly lower change in predicted biological age in treated patients. The 30 mg twice-daily arm produced the most consistent signal with nine significant comparisons, ahead of 60 mg once daily with seven and 30 mg once daily with five. A permutation test with patient-level label shuffling put the chance expectation at 0.15 significant comparisons.^[3]

The paper's explicit year figures are narrower than the press coverage suggests. At week 4 the 60 mg once-daily group showed reductions of 2.71 to 3.46 years across all four chronological clocks, with no significant change on either mortality clock. Organ-specific mortality variants of OrganAge produced the largest numbers, with the artery clock reading 6.95 to 16.57 years lower than placebo across all timepoints and treated arms.^[3] Insilico's release summarizes the peak as "approximately 3-4 years reversal in biological age and up to 6 years in a certain aging clock" at week 4 in the 30 mg twice-daily group.^[20] The paper's own strongest cross-clock agreement is indeed at week 4 in that arm, but the 3-to-4-year range it prints belongs to the 60 mg once-daily arm, and the figure closest to six years is an organ-specific clock rather than one of the six headline models. Twelve weeks is the length of the sampling window, not the time to the reported peak.

By week 12 fewer comparisons were significant. Direct week-4 to week-12 comparison showed no significant further shift in any arm-clock pair, which the authors describe as a plateau rather than a rebound, and they note that the underlying proteomic changes mostly continued to develop rather than fade: only 5 to 9 percent of protein shifts in the 30 mg twice-daily and 60 mg once-daily arms were transient.^[3]

Separating aging from the treated lung

The authors address the obvious confound directly and do not claim to have resolved it. Their argument is one of dissociation. The arm with the largest lung-function gain in the original trial, 60 mg once daily, was not the arm with the most consistent aging-clock response; change in forced vital capacity explained little of the variance in predicted biological age across the six clocks (median R-squared 0.06, range 0.01 to 0.18, n=42). Compared against age-associated protein trajectories in 55,319 older UK Biobank participants, the 30 mg twice-daily regimen's protein changes ran opposite to normal aging (Spearman's r = -0.30, P < 0.01) while the 60 mg once-daily regimen showed no such correlation (r = -0.097, P = 0.37). Proteins uniquely affected by 30 mg twice daily were enriched for pentose phosphate, glutathione and cholesterol metabolism, whereas those unique to 60 mg once daily were enriched for WNT signaling and immune pathways closer to the fibrotic mechanism.^[3]

Gene-set enrichment analysis found opposite senescence trajectories in placebo and treated groups. The SenMayo senescence signature rose in placebo and fell in every treated arm, with EREG, ESM1, IGFBP4, ITGA2, MMP10, MMP13 and SPP1 appearing in the leading edge of all three. Growth-factor signaling pathways involving RTK, RAS, ERK and PI3K were downregulated in all treated arms, and six IGF-binding proteins and related proteins fell relative to baseline across regimens.^[3]

The paper's stated limitations are the small sample, the short duration, the reliance on computational rather than experimental readouts, and the absence of other omics layers, which together "precluded clear deconvolution of anti-fibrotic and anti-aging effects." The authors say the question requires studies in non-IPF populations where severe fibrosis cannot confound interpretation.^[3] Michael Levitt, the 2013 Nobel laureate in chemistry, is quoted in Insilico's release making the same point: "This trial cannot yet separate slower aging from a treated lung, and the authors say so plainly. The experiment in healthy volunteers is the one I want to see next."^[20]

Insilico said the results would be presented by Zhavoronkov at the Nature conference "Redefining Healthcare in the Age of AI," held at the Sorbonne in Paris from 8 to 10 September 2026.^[20][21]

Data and code

The Olink dataset is deposited with the China National Center for Bioinformation as OMIX008341 under BioProject PRJCA033225, submitted in December 2024 and released in March 2025. Access is controlled, so the data are available on request rather than by download.^[22] The clock implementations were standardized in a Python library, proteoclock, published by Insilico under an MIT license.^[3][23] ProtAge was accessed through its public repository with model weights obtained by direct collaboration.^[3]

Regulatory status

Rentosertib remains investigational and has not been approved by any regulatory authority.^[6] Insilico reports that the US Food and Drug Administration granted it orphan drug designation for IPF in February 2023, and that China's Center for Drug Evaluation granted it breakthrough therapy designation in May 2025.^[6][10][11] Both designations are procedural: they affect development and review support, not any conclusion about efficacy.

The name itself is a deliberate reference. Zhavoronkov said in the 2026 release that he was "happy that the drug carries the name of the true hero of the AI drug discovery revolution, our co-CEO, Dr. Ren," pointing at co-chief executive Feng Ren.^[20]

How the "first AI drug" claim is used

The superlatives attached to rentosertib are almost all Insilico's own, and they have shifted as the program advanced. In 2023 the company and outlets including CNBC described it as the first fully AI-generated drug to enter clinical trials in patients.^[24] The 2026 aging-clock release calls the work the "first-in-class evaluation of an AI-driven candidate," the "first universal consensus across six independently-developed proteomic aging clocks," and the "first scalable blueprint for dual-purpose clinical trials."^[20] The Phase 3 release calls GENESIS-IPF-3 "the world's first Phase III trial of a generative AI-driven innovative drug."^[6]

What is documented in the peer-reviewed record is narrower and still substantial: a target that the earlier literature had studied mainly in oncology rather than fibrosis, a molecule generated by a generative-chemistry platform, 18 months from target discovery to candidate nomination, a completed Phase 2a trial that met its safety endpoint and produced a dose-dependent lung-function trend in small arms, and an exploratory biomarker analysis whose authors decline to claim that aging itself was modulated.^[1][2][3] Whether the compound works is a question the 320-patient Phase 3 trial is designed to answer, with primary completion estimated for October 2029.^[5]

References

1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15Ren, F., Aliper, A., Chen, J., et al. "A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models." *Nature Biotechnology* 43, 63-75 (2025), published online 8 March 2024. nature.com/...s41587-024-02143-0
2. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10Xu, Z., Ren, F., Wang, P., et al. "A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial." *Nature Medicine* 31, 2602-2610 (2025), published 3 June 2025. nature.com/...s41591-025-03743-2
3. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18Zhavoronkov, A., Galkin, F., Chen, S., et al. "Integration of proteomic aging clocks in a phase 2a clinical trial supports simultaneous geroprotective assessment." *Nature Biotechnology*, published 7 September 2026. nature.com/...s41587-026-03286-y
4. ^Drug Target Review. "First AI-designed drug, Rentosertib, officially named by USAN." 14 March 2025. drugtargetreview.com/...-rentosertib-named-by-usan
5. ^1 ^2 ^3 ^4 ^5 ^6ClinicalTrials.gov. "A Prospective Phase 3, Randomized, Double-blind, Placebo-controlled, Parallel-group Study Evaluating the Efficacy and Safety of Rentosertib (INS018_055) Administered Orally Over 52 Weeks in Patients With Idiopathic Pulmonary Fibrosis (IPF)." NCT07687459. clinicaltrials.gov/...NCT07687459
6. ^1 ^2 ^3 ^4 ^5 ^6 ^7Insilico Medicine. "Insilico Medicine Doses First Patient in GENESIS-IPF-3, the World's First Phase III Trial of a Generative AI-Driven Innovative Drug." Press release, 10 September 2026. insilico.com/...-doses-first-patient-genesis-ipf-3
7. ^PubChem. "Rentosertib," CID 164938183. National Center for Biotechnology Information. pubchem.ncbi.nlm.nih.gov/...164938183
8. ^Aladinskiy, V., Kruse, C., Qin, L., et al. "Discovery of Bis-imidazolecarboxamide Derivatives as Novel, Potent, and Selective TNIK Inhibitors for the Treatment of Idiopathic Pulmonary Fibrosis." *Journal of Medicinal Chemistry* 67, 19121-19142 (2024). doi.org/...acs.jmedchem.4c01580
9. ^Tang, Q., Xiao, D., Veviorskiy, A., et al. "AI-Driven Robotics Laboratory Identifies Pharmacological TNIK Inhibition as a Potent Senomorphic Agent." *Aging and Disease* 17, 432-451. doi.org/...AD.2024.1492
10. ^1 ^2 ^3Insilico Medicine. "Insilico's Rentosertib Inhalation Solution Receives IND Clearance for the World's First AI-Driven Candidate to Enter Direct-to-Lung Clinical Study." Press release, 28 April 2026. insilico.com/...cos-rentosertib-inhalation-solutio
11. ^1 ^2Insilico Medicine. "Insilico Medicine receives FDA Orphan Drug Designation for generative AI discovered and designed drug for idiopathic pulmonary fibrosis." Press release via EurekAlert, February 2023. eurekalert.org/...979045
12. ^1 ^2ClinicalTrials.gov. "Study Evaluating INS018_055 Administered Orally to Subjects With Idiopathic Pulmonary Fibrosis (IPF)." NCT05938920. clinicaltrials.gov/...NCT05938920
13. ^ClinicalTrials.gov. "A Phase IIa, Randomized, Double-Blind, Placebo-Controlled Study Evaluating the Safety, Tolerability, Pharmacokinetics, and Efficacy of INS018_055 Administered Orally to Subjects With Idiopathic Pulmonary Fibrosis (IPF)." NCT05975983. clinicaltrials.gov/...NCT05975983
14. ^1 ^2 ^3Insilico Medicine. "Insilico Medicine Announces Nature Medicine Publication of Phase IIa Results Evaluating Rentosertib." Press release, June 2025. prnewswire.com/...pioneering-ai-approach-302472070
15. ^Argentieri, M. A., et al. "Proteomic aging clock predicts mortality and risk of common age-related diseases in diverse populations." *Nature Medicine* 30, 2450-2460 (2024). doi.org/...s41591-024-03164-7
16. ^1 ^2Goeminne, L. J. E., et al. "Plasma protein-based organ-specific aging and mortality models unveil diseases as accelerated aging of organismal systems." *Cell Metabolism* 37, 205-222 (2025). doi.org/...j.cmet.2024.10.005
17. ^Kuo, C.-L., et al. "Proteomic aging clock (PAC) predicts age-related outcomes in middle-aged and older adults." *Aging Cell* 23, e14195 (2024). doi.org/...acel.14195
18. ^Galkin, F., Chen, S., Aliper, A., Zhavoronkov, A. & Ren, F. "AI-driven toolset for IPF and aging research associates lung fibrosis with accelerated aging." *Aging* 17, 1999-2014 (2025). doi.org/...aging.206295
19. ^Xu, H., Chen, J., Chen, D., Mao, K. & Han, J.-D. J. "Proteome-aware organ proxy aging clocks." bioRxiv preprint, 2026. doi.org/...2026.04.24.720503
20. ^1 ^2 ^3 ^4 ^5 ^6 ^7Insilico Medicine. "Nature Biotechnology | Insilico's AI-Driven IPF Candidate Rentosertib Shows Potential for Biological Age Reversal, as Assessed by Six Proteomic Aging Clocks." Press release, 7 September 2026. insilico.com/...rentosertib-proteomic-aging-clocks
21. ^Nature Careers. "Redefining Healthcare in the Age of AI," Sorbonne, Paris, 8-10 September 2026. nature.com/...redefining-healthcare-in-the-age-of-ai
22. ^China National Center for Bioinformation, OMIX database. "Proteomic for IPF patients plasma samples in INS018_055 phase 2a," accession OMIX008341. ngdc.cncb.ac.cn/...OMIX008341
23. ^Insilico-org/proteoclock. "Package for predicting biological age from OLINK profiles." GitHub, MIT license. github.com/...proteoclock
24. ^Field, H. "The first fully A.I.-generated drug enters clinical trials in human patients." CNBC, 29 June 2023. cnbc.com/...gins-clinical-trials-in-human-patients