Virtual Biotech AI System Analyzes Clinical Trials for Drug Discovery

Researchers have developed a multi-agent artificial intelligence framework called the Virtual Biotech, designed to assist with drug discovery by analyzing large-scale biomedical data and clinical trial results.

The Virtual Biotech system consists of up to 37,000 AI scientist agents, each coordinated by a virtual chief scientific officer. This platform assigns individual agents to analyze specific clinical trials, enabling the review of outcomes from more than 55,000 trials across a range of therapeutic areas.

To carry out its analyses, the Virtual Biotech integrates a variety of biomedical data sources. These include single-cell gene expression profiles, protein-protein interaction networks, pharmacogenomic variants, and gene expression changes resulting from drug treatments. The system is designed to produce reproducible outputs, such as code, figures, and data files, which can be audited by human scientists.

In its research, the Virtual Biotech identified that drugs targeting cell-type-specific genes were about 48% more likely to reach the market and had approximately 32% fewer adverse events compared to other approaches. The platform also proposed new therapeutic strategies, such as an antibody-drug conjugate targeting CD276 for lung cancer, based on its analysis of existing data.

What the numbers show

• 37,000 AI scientist agents were used in the Virtual Biotech system.
• More than 55,000 clinical trials were analyzed by dedicated agents.
• Drugs targeting cell-type-specific genes were 48% more likely to reach market.
• These drugs showed 32% fewer adverse events.
• The peer-reviewed paper was published on 17 September 2026.

The Virtual Biotech framework has also been applied to case studies, such as a failed ulcerative colitis trial targeting OSMRβ. In this example, the system suggested biomarker-guided enrollment strategies that could potentially inform future trial designs.

According to the published research, the platform enables transparent review of agent reasoning processes. Human scientists can examine the steps taken by the AI agents, as well as the data and code generated during their analyses.

The peer-reviewed findings describing the Virtual Biotech system and its applications were published in the journal Science on 17 September 2026. This publication outlines the technical details of the multi-agent approach and summarizes the outcomes of its analyses.

By coordinating thousands of AI agents and integrating diverse data sources, the Virtual Biotech system represents a procedural advancement in the use of artificial intelligence for drug discovery research. The reproducibility and auditability of its outputs allow for ongoing evaluation and refinement by the scientific community.

* This article is based on publicly available information at the time of writing.