ANTARES project launches AI push to find new antibiotics

A German public-private partnership called ANTARES is using AI and lab testing to speed the search for new antibacterials against drug-resistant infections. The three-year project began July 1, 2026, and aims to improve discovery rates while publishing its data and models for wider use.

Why it matters: - ANTARES is targeting the urgent need for new antibiotics as antimicrobial resistance erodes the value of existing treatments. - The project is designed to shorten early-stage drug discovery and improve the odds of finding active molecules. - The consortium plans to release models and standardized datasets in FAIR format, which could help other antibiotic research efforts.

What happened: - ANTARES launched as a public-private partnership funded by the German Federal Ministry of Research, Technology and Space. - The project brings together Fraunhofer ITMP in Hamburg, the Helmholtz Centre for Infection Research in Braunschweig and Enamine Deutschland GmbH in Frankfurt am Main. - The project started July 1, 2026, and is set to run for three years. - The funding totals 1.18 million euros. - The grant is managed by Project Management Jülich under grant number 03LWH0186A.

The details: - The consortium will use AI models to identify novel antibiotics from publicly available datasets first. - The team will then refine the models through iterative rounds of high-throughput testing on substances predicted to have antibiotic activity. - Structurally similar compounds predicted to be inactive will also be tested to improve the model’s ability to separate active from inactive molecules. - Fraunhofer ITMP’s Bernhard Ellinger said the iterative cycle of prediction and experimental validation is a key advantage because some programs rely on only one training round. - Enamine will contribute to the design and creation of small molecules for antibiotic discovery at wider scale. - The project will use explainable AI methods to extract structure-activity relationships from the models. - A generative model will propose new structures with antibiotic activity. - The models and standardized datasets will be published FAIR to support further antibiotic discovery. - HZI’s Mark Brönstrup said the goal is to save time and improve success rates at the discovery stage by combining AI with high-throughput experimentation. - The project will also describe the mechanism of action and activity range of identified molecules. - The consortium can make and test newly synthesized libraries, not just known datasets.

Between the lines: - ANTARES is trying to solve a classic bottleneck in antibiotic development: promising science often fails before it becomes a usable drug. - The partnership structure matters because it connects academic screening, infection biology and compound generation in one workflow. - Enamine described the field as a “broken market,” reflecting the weak commercial incentive for antibiotic development. - The project’s focus on public datasets, explainability and FAIR release suggests an effort to make the results reusable beyond the immediate partnership.

What's next: - The partners will train, test and iteratively improve the AI models using experimental feedback. - The consortium will expand from known datasets into newly synthesized chemical libraries. - Future work will aim to identify active molecules, define how they work and map the range of infections they may address. - The project’s output is expected to include reusable datasets and models that can support follow-on antibiotic research.

The bottom line: - ANTARES is betting that tighter AI-lab loops can make antibiotic discovery faster, cheaper and more productive at a time when new treatments are badly needed.