Whitepaper
AI in drug discovery: what the evidence shows
AI can decrease the time for some steps in drug discovery. But a fast result in research is not proof that a drug is safe and effective. This whitepaper is for pharmaceutical enterprises and healthcare scaleups that must make decisions about AI. It is not medical advice.
Where AI can help
- Target identification: AI can examine large biological datasets and find possible disease targets.
- Molecule design: Generative models can suggest new molecules with specified properties.
- Toxicity prediction: Models can identify possible safety problems before laboratory tests start.
- Trial design: AI can help teams to select patient groups and doses from existing data.
Each of these outputs is a prediction. Scientists must examine each prediction, and laboratory and clinical tests must confirm it.
What the evidence shows
Rentosertib is a drug candidate for idiopathic pulmonary fibrosis. Insilico Medicine found the target and designed the molecule with generative AI. Nature Medicine published the results of a phase 2a trial in June 2025.
The trial had 71 patients at 22 sites in China, and treatment continued for 12 weeks. The primary endpoint was safety. Adverse event rates were similar in all groups. In the 60 mg once-daily group, the mean change in forced vital capacity was +98.4 mL. In the placebo group, the mean change was -20.3 mL.
This result is positive, but the trial was small and short. It does not show that the drug is effective in a larger population. Larger trials must supply that evidence.
What to prepare
The model is only one part of the work. Before you use AI in a regulated workflow, prepare these items:
- Data readiness: Examine the quality, source, and permissions of each dataset.
- Validation: Set the tests that a model must pass before people use its results.
- Records: Record the data, model version, and person that approved each result.
- Data protection: Identify the personal data and apply the applicable law, such as the GDPR.
- Integration: Connect the model to the systems that teams already use.
Orchestron works in life sciences, with patient facing services, records integration, and applied AI.
Sources
Get the full whitepaper
This page is a summary. Contact us to ask for the full whitepaper.