Developing drug molecules with AI: the case of Rentosertib
A precedent has been established in clinical pharmacology: the small molecule Rentosertib (Rentasertib / INS018_055), fully generated by artificial intelligence algorithms from the identification of a biological target through the synthesis of the finished compound, demonstrated an unforeseen systemic effect of reducing patients’ predicted biological age.
The drug was developed by the biotechnology company Insilico Medicine to treat idiopathic pulmonary fibrosis—a progressive, incurable disease in which alveolar lung tissue is replaced by fibrotic scar tissue. The entire early-development cycle, from identifying the protein target to obtaining a candidate for preclinical trials, took about 18 months, which is three times faster than the traditional industry cycle of 4 to 6 years.
The TNIK biological target: a dual role in fibrosis and cellular aging
The starting point for creating the compound was an analysis of multi-omics databases. The AI platform identified the therapeutic significance of the TNIK kinase (Traf2- and Nck-interacting kinase). This protein regulates key signaling pathways, including the Wnt/beta-catenin cascade, that activate fibroblasts during pathological fibrogenesis.
At the same time, the TNIK kinase participates in molecular pathways associated with cellular senescence and chronic low-grade inflammation. Inhibition of the protein demonstrated the ability to block the fibrotic transformation of the lung epithelium while also normalizing signaling pathways in the extracellular matrix.
Results of 2a-phase clinical trials and measurements of proteomic age
More than 40 patients with confirmed pulmonary fibrosis participated in the placebo-controlled 2a-phase study. The primary clinical endpoint was achieved: respiratory function in the active-treatment group stabilized and showed positive dynamics.
A secondary analysis of blood plasma samples from participants in the study group revealed systemic changes in the protein profile. For an objective assessment, six independent machine-learning models (“proteomic aging clocks”) developed by research teams at Harvard, Oxford, and Peking University were used:
| Biological age assessment model | Developer / Data source | Recorded reduction in biological age (4 weeks) |
|---|---|---|
| Harvard proteomic panel | Harvard Medical School | Average of -3.2 years |
| Oxford plasma clock | Oxford University Proteomics | Average of -3.6 years |
| Peking epigenetic model | Beijing Genomics Institute | Average of -4.1 years |
| Multifactor deep-learning model | Longevity Consortium | Maximum measurement of up to -5.8 years |
All six independent algorithms recorded a statistically significant shift in the concentrations of circulating proteins toward values characteristic of younger cohorts. A control comparison with a dataset of 55 000 UK Biobank profiles confirmed that the changes affected precisely those protein fractions whose levels monotonically increase or decrease with chronological aging.
The dosing paradox: why rejuvenation cannot be reduced to improved lung function
The reviewers’ initial hypothesis was that the algorithms had mistaken improved well-being in severely ill patients for true biological rejuvenation—the reduction of systemic stress in the body as breathing became easier. However, a detailed analysis of the dosing regimens disproved a direct linear correlation:
- Dosage of 60 mg once daily: demonstrated maximum efficacy in treating pulmonary fibrosis (the best spirometry dynamics), while the change in estimated biological age was moderate (about 1.5–2 years);
- Dosage of 30 mg twice daily: produced the maximum shift in aging markers (a reduction of 3.5–5.8 years), despite comparable or slightly lower therapeutic improvement in lung tissue.
This discrepancy indicates an independent pharmacodynamic mechanism by which the molecule affects general metabolism and protein homeostasis pathways.
Study limitations and the perspective of the scientific community
Nobel laureate Michael Levitt and leading longevity researchers, including Vadim Gladyshev and Eric Topol, call for a strict distinction between the terms:
- What has been proven: the Rentosertib compound reversibly changes the concentration of specific age-dependent blood plasma proteins in patients with pulmonary pathology during a 4-week course;
- What has not been proven: there is no direct evidence of an actual extension of lifespan, restoration of the morphology of unaffected organs, or reversal of degenerative processes in healthy people;
- Necessary step: conducting studies in a healthy volunteer population to test the hypothesis of a geroprotective effect in the absence of an underlying chronic inflammatory process.
Compare models before you start
The service sets its plans, limits and model catalog. If they differ from this article, contact us so we can update it and record a new review date.
Browse modelsAffiliate link: your price stays the same and the project earns a commission.