Insilico Medicine is advancing rentosertib into Phase III trials for idiopathic pulmonary fibrosis, moving an AI-originated programme into the late-stage test that determines whether promising early results translate into meaningful patient benefit.
IPF progressively scars lung tissue and carries a median survival of only two to four years after diagnosis. Rentosertib is an oral inhibitor of TNIK, a target selected through Insilico’s computational biology platform rather than conventional starting assumptions.
AI connected disease biology to a new target and molecule
PandaOmics analysed genomics, trials, literature and patents to identify TNIK as a node connecting fibrosis, inflammation and ageing pathways. The approach looked beyond receptor tyrosine kinases targeted by existing antifibrotic drugs.
Chemistry42 then generated molecules designed for the target pocket while balancing pharmacological properties. Insilico physically synthesised 79 candidates and advanced the 55th iteration, reaching preclinical nomination in 18 months.
That history shows more than high-throughput screening. AI influenced target choice and created new chemistry, while laboratory structure and biological work validated that the proposed interaction existed.
Peer-reviewed publications document target prioritisation, chemistry, preclinical evidence and early human studies. This traceability is important because the field needs inspectable programmes rather than claims of an “AI drug” based on one assisted step.
Phase IIa data justify a larger test, not a conclusion
A randomised trial involved 71 patients at 22 Chinese sites over 12 weeks. The 60 mg group showed a mean forced vital capacity gain of 98.4 mL, compared with a 20.3 mL decline for placebo, with reportedly manageable adverse events.
The signal is encouraging, but the cohort and duration are small for a chronic disease. Phase III must test more diverse patients, longer treatment and clinically meaningful outcomes while characterising less common safety problems.
The FDA granted orphan-drug designation in 2023, which recognises the rare-disease programme and offers development incentives; it is not an approval or efficacy endorsement.
Proteomic clocks and senescence markers provide exploratory evidence about biological effects. They should complement established endpoints rather than substitute surrogate sophistication for how patients feel, function and survive.
Trial design should account for background therapy, disease severity and geographic diversity. The early study was conducted in China, so a larger programme can show whether the signal generalises across health systems and populations. Public registration, prespecified endpoints and complete reporting will make the eventual result more credible.
Phase III tests the whole discovery thesis against reality
AI drug discovery has often been judged by speed to candidate. Rentosertib now faces the same decisive standard as every medicine: a controlled trial capable of showing benefit and acceptable risk.
A successful result would validate this programme, not every algorithmic target or generated molecule. Failure would likewise require analysis of trial design, target biology, molecule and patient selection rather than a blanket verdict on AI.
Insilico should preserve complete provenance from source data through model versions and experimental decisions. Regulators and partners need to reconstruct how the candidate emerged and why changes were made.
Commercial pressure must not outrun clinical evidence. Patients with a severe disease deserve access to promising trials alongside clear communication about uncertainty.
Rentosertib is important because it has crossed from computational promise to human data. Phase III will be expensive, slow and resistant to storytelling—the qualities that make it a credible test.
If the drug succeeds, AI will have helped originate new biology and chemistry that improved care. Until then, the achievement is narrower but real: an algorithmically initiated hypothesis has survived enough scientific scrutiny to earn the field’s most demanding clinical examination.
Patients, clinicians and regulators will judge outcomes, adverse events and durability rather than discovery speed.
