Insilico Medicine is taking its generative AI drug-discovery strategy into a new phase of testing: not just whether an AI-designed drug can treat disease, but whether it can also shift molecular markers associated with biological aging. A study published in Nature Biotechnology reports that rentosertib, an investigational idiopathic pulmonary fibrosis (IPF) drug whose target and molecule were developed using AI, produced lower predicted biological-age scores across six independent proteomic aging clocks in a Phase IIa dataset. The finding could give drug developers a new way to investigate geroprotective effects inside conventional disease trials, although it does not yet demonstrate that the drug slows aging or extends human lifespan.

The most interesting part of Insilico Medicine’s latest AI drug-discovery story is not simply that an algorithm helped design a drug. It is that the company is now using AI-derived biological-age models to interrogate what happens after that drug reaches patients.

The study, published September 7 in Nature Biotechnology, analyzed longitudinal serum proteomics from a Phase IIa clinical trial of rentosertib in patients with idiopathic pulmonary fibrosis. Researchers applied six independently developed proteomic aging clocks — including ProtAge, OrganAge, PAC, ipfP3GPT and PAOPAC — to samples from the trial.

All six models showed lower predicted biological age in rentosertib-treated groups compared with placebo, with the strongest and most consistent signal appearing around Week 4 in the 30 mg twice-daily regimen.

That convergence matters because biological-age clocks are not interchangeable. They are statistical models trained on molecular or physiological data to estimate aspects of aging that chronological age alone cannot capture. Recent research has shown that proteomic clocks can provide information associated with disease risk and mortality, but researchers are still working to understand how reliably these models measure intervention-driven changes in aging.

In other words, the Insilico study is potentially important for AI-enabled drug development, but it is not yet evidence that rentosertib is a clinically proven longevity drug.

From AI target discovery to human biology

Rentosertib, formerly known as ISM001-055, is an oral small-molecule inhibitor of TNIK, or TRAF2- and NCK-interacting kinase.

Insilico’s drug-discovery platform was used to identify TNIK as a target associated with fibrosis and aging biology, after which its generative chemistry system, Chemistry42, was used to design and optimize the molecule. The program subsequently advanced into human clinical development and has now entered Phase III testing for IPF.

That progression makes rentosertib an unusually useful case study for the broader AI-drug-discovery market.

The traditional promise of generative AI in pharmaceuticals has been largely computational: search larger chemical spaces, identify potential targets, design molecules and shorten early discovery cycles.

Insilico is attempting to extend that model beyond molecule generation.

Its latest study effectively asks whether AI can connect target biology, molecular design, clinical development and aging biomarkers into one continuous development loop.

That is a much more ambitious proposition.

What the aging clocks actually found

The researchers analyzed 2,841 proteins from serum samples collected during the 12-week Phase IIa study. Across the six clocks, treated patients generally showed reductions in predicted biological age, while placebo participants showed minimal changes or slight increases. The researchers reported 21 statistically significant treatment-versus-placebo comparisons across the different clocks and time points, with the strongest concentration at Week 4.

The analysis also found evidence that rentosertib altered proteins associated with cellular senescence, fibrosis, extracellular-matrix remodeling and growth-factor signaling.

A comparison against proteomic data from 55,319 UK Biobank participants provided another layer of analysis. The 30 mg twice-daily regimen showed a significant negative correlation between treatment-induced protein changes and normal age-associated trajectories, suggesting that some of the molecular changes moved in the opposite direction from patterns normally associated with aging.

But there is an important caveat.

The researchers explicitly note that proteomic aging clocks alone cannot separate aging effects from disease-related effects. A drug that improves fibrosis could also change proteins associated with aging without necessarily altering the underlying aging process.

That distinction is crucial for enterprise AI and pharmaceutical R&D teams evaluating these technologies.

Why this could change clinical-trial design

The bigger opportunity may therefore be the trial framework, rather than rentosertib itself.

Most drug trials are designed around a particular disease, endpoint and treatment hypothesis. Aging is usually treated as background biology rather than a measurable therapeutic endpoint.

Insilico’s approach reverses that assumption.

Researchers prospectively collected proteomic samples during the IPF trial, allowing the same clinical study to generate data about both disease response and potential geroprotective effects.

That could eventually enable pharmaceutical companies to screen promising therapies for broader aging-related activity without immediately launching separate, lengthy longevity trials.

The idea is gaining traction across longevity research. A 2025 npj Aging paper recommended systematic collection of aging biomarkers in clinical trials to improve participant stratification, intervention assessment and validation of aging biomarkers.

The regulatory challenge, however, remains significant.

The U.S. Food and Drug Administration’s Biomarker Qualification Program allows developers to establish biomarkers for defined contexts of use, but a biomarker must demonstrate that it can reliably support a specific interpretation in drug development. FDA also makes clear that a biomarker does not have to be formally qualified to be used within an individual drug-development program.

For AI-driven longevity research, that distinction could become increasingly important.

A model can produce a statistically compelling biological-age prediction. Turning that prediction into an accepted clinical endpoint is a much higher bar.

AI drug discovery moves toward closed-loop biology

The rentosertib program illustrates where AI-enabled pharmaceutical R&D could be heading.

Instead of treating AI as a single tool for molecule generation, companies are building interconnected systems covering target discovery, molecular design, biomarker analysis and clinical decision-making.

That creates a potential closed-loop drug-discovery architecture: AI identifies biological hypotheses, generative models design candidates, clinical trials produce molecular and physiological data, and machine-learning models analyze those outcomes to refine the next generation of therapeutic hypotheses.

Insilico’s commercial strategy reflects that broader platform ambition. The company has also been developing foundation-model benchmarks and AI systems aimed at expanding the computational capabilities available across drug discovery.

The industry is increasingly crowded, however. AI-enabled pharmaceutical development now includes specialized biotech companies, pharmaceutical R&D groups and technology providers building models for proteins, molecules, biological pathways and clinical data.

For Insilico, the differentiator will not simply be whether an AI model can generate a molecule. It will be whether the resulting programs can repeatedly produce clinically meaningful outcomes faster, more efficiently and with stronger biological validation.

Rentosertib remains investigational. Its Phase III program will ultimately provide a much stronger test of whether the earlier clinical signals translate into durable benefit for people with IPF.

The biological-age findings add another dimension to that experiment.

They suggest that clinical trials could become not only tests of whether a drug treats a disease, but also data-generating environments for understanding how therapies interact with the biology of aging.

If that model proves reproducible across drugs and diseases, AI could become useful not just for discovering medicines — but for helping researchers determine which biological processes those medicines are actually changing in humans.

Market Landscape

The AI drug-discovery market is moving from isolated computational tools toward integrated platforms combining target identification, generative chemistry, proteomics, foundation models and clinical biomarkers.

Proteomic aging clocks are particularly relevant because they translate thousands of circulating proteins into an estimate of biological age. Research using large datasets such as the UK Biobank has shown that proteomic aging measures can correlate with multiple chronic diseases and mortality risk.

The next challenge is intervention validation.

Aging clocks were initially valuable largely as observational tools. The more difficult question is whether a treatment-induced change in a clock represents a meaningful improvement in health rather than simply a change in the molecular signature being measured.

That makes Insilico’s approach strategically interesting but still experimental.

For pharmaceutical R&D organizations, the emerging opportunity is to combine AI drug design + multi-omics + longitudinal biomarkers + clinical trials into a single evidence-generation system.

Top Insights

• Insilico’s rentosertib study links AI-designed drug development with biological-age measurement, giving pharmaceutical teams a potential framework for evaluating geroprotective effects during disease trials.
• Six independent proteomic aging clocks showed consistent age-score reductions, strengthening the signal while leaving open whether molecular age reversal translates into meaningful clinical outcomes.
• UK Biobank comparisons showed treatment-induced proteins moving against age-associated trajectories, providing additional evidence that rentosertib affects pathways linked to aging biology.
• The approach could change clinical-trial design, allowing biotech and pharmaceutical teams to collect aging biomarkers alongside conventional disease endpoints rather than running separate longevity studies.
• Regulatory validation remains the major hurdle, because AI-generated biological-age signals must demonstrate a reliable clinical context of use before becoming established drug-development endpoints.

Power Tomorrow’s Intelligence — Build It with TechEdgeAI