MMOOZZGG.com

Areas

AI drug discovery: find a candidate faster, not skip the trials

2 min read7 sources

AI drug discovery has shown a measurable speedup at the step of finding a molecule, in Russia and in the world. Below are cases from pharma companies and institutes, and the first clinical proof. And one caveat: AI speeds up the search, while trials and registration stay as they were.

On this page · 4 sections

Russia

A pharma company, together with the AI lab of a bank, cut the step of designing antibody structures from about 3 years to 2 months. The lab confirmed the hypotheses, and the cycle from idea to validated structure became 3 times shorter (2024). Another company built a pipeline: the lab stage of molecule analysis takes weeks instead of 7–14 months. The company had 78 registered drugs at the end of 2024 and 32% more clinical trials. A third company, by its own estimate, halved development time and raised the share of successful candidates at least 2.5 times. The Kurchatov Institute found more than 100 genomic loci linked to stroke.

The world

A candidate molecule against pulmonary fibrosis, found with generative AI, passed phase IIa on 71 patients: at the highest dose, lung function rose by 98.4 ml on average, against a fall of 20.3 ml on placebo. According to the company, the path to the candidate took about 18 months instead of the usual 4.5 years (Nature Medicine, 2025). An open model for protein structure prediction: more than 200 million structures, more than 3 million researchers in 190 countries, the 2024 Nobel Prize, and more than 30% of the papers that use it deal with diseases.

The caveat

All the cases are about candidates and the steps before the clinic. A molecule becomes a drug only after clinical trials approved by Russia’s Ministry of Health (drug law 61-FZ) and registration; advertising of unregistered drugs is not allowed. Our area is called “molecule search”, not “drug creation”, and this page keeps the same line.

What we do

Reviews with sources. Computational experiments on open databases are in preparation: they will come out as research papers with methods and limits. We are looking for partners for lab testing. More on the research page.

Questions and answers

Are there already drugs created by AI?

The sources on this page speak about candidates in clinical trials, including one that passed phase IIa. A candidate becomes a drug only after registration.

Can an ad say “a drug created by AI”?

No. Before registration it is a compound or a candidate, and advertising of unregistered drugs is banned.

Sources and fact-checking

Checked by the editors on 29 September 2026.

  1. Digital Platforms, case no. 2495, 2024: R-Pharm (platforms.su)
  2. Digital Platforms, case no. 3796, 2 September 2026: RTU MIREA (platforms.su)
  3. Digital Platforms, case no. 2496: BIOCAD (platforms.su)
  4. Digital Platforms, case no. 2510: Generium (platforms.su)
  5. Pharmprom.news, 4 June 2025: Insilico Medicine, USA and Hong Kong (pharmprom.news)
  6. Digital Platforms, case no. 2586: Kurchatov Institute (platforms.su)
  7. Google DeepMind, 25 November 2025: Google DeepMind (deepmind.google)

Found a mistake? Tell us

Publications

Cases on this topic: 5

  • Market case · Russia

    R-Pharm a candidate molecule structure in about 2 months instead of 3 years

    The pharma company, together with a bank’s AI lab and the AIRI institute, used generative models for AI drug design of antibodies against the CD137 receptor. The AI proposed dozens of candidate molecules, and the lab confirmed the hypotheses; according to the catalog, the structure design stage fell from about 3 years to about 2 months.

    Drug discovery and pharma

    2024 · source: platforms.su

    Read the case
  • Market case · Russia

    RTU MIREA a 16–17% error in drug demand forecasts one month ahead

    The university built a pilot hybrid model for pharmaceutical demand forecasting in the Russian market, with its sharp price changes and short sales histories of new generic drugs. Gradient boosting ensembles gave a 16–17% error by the WMAPE metric over a one-month horizon.

    Drug discovery and pharma

    2 September 2026 · source: platforms.su

    Read the case
  • Market case · Russia

    BIOCAD lab stage of molecule analysis in weeks instead of 7–14 months, 78 drugs registered by the end of 2024

    BIOCAD is one of the first pharmaceutical companies in Russia to apply machine learning to molecule design: its platform combines screening, gene synthesis and mathematical modeling of structures. The lab stage of molecule analysis fell from 7–14 months to a few weeks, and on some steps to days.

    Drug discovery and pharma

    source: platforms.su

    Read the case
  • Market case · Russia

    Generium according to the company, drug development takes half the time, and the share of successful candidates is at least 2.5 times higher

    The company uses machine learning at the key stages: target discovery, virtual screening of molecule libraries, forecasts of pharmacokinetics and toxicity, and improvement of synthesis and delivery systems. According to the company’s vice president, development time fell by half, and the share of successful candidates rose at least 2.5 times.

    Drug discovery and pharma

    source: platforms.su

    Read the case
  • Market case · World

    Insilico Medicine, USA and Hong Kong a molecule found by generative AI passed phase IIa: lung function +98.4 mL vs -20.3 mL on placebo

    Rentosertib, a drug candidate for idiopathic pulmonary fibrosis found by generative AI, passed phase IIa with 71 patients in 22 centers. On the top dose, forced vital capacity rose by 98.4 mL on average, while on placebo it fell by 20.3 mL. The result appeared in Nature Medicine on 3 June 2025.

    Drug discovery and pharma

    4 June 2025 · source: pharmprom.news

    Read the case
All cases: 62
Contact us