Researchers introduced LabAgent for automating scientific laboratories

A group of scientists developed LabAgent — a specialized framework designed to preserve and reproduce research methodologies in scientific laboratories. The system solves the problem of losing expertise when personnel changes, ensuring continuity of experiments and automation of routine tasks.

The system's operation is based on two key mechanisms: verification of skill execution and logging of corrective actions. This allows the agent not only to repeat previously performed protocols but also to consider accumulated experience to prevent typical errors made by previous researchers. The framework is distributed under the CC BY-NC-SA 4.0 license.

The effectiveness of the solution was tested in four areas: prediction of drug properties, analysis of biomedical data, prediction of protein variant effects, and statistical genetics. During the tests, LabAgent showed better results compared to commercial universal AI agents in all mentioned disciplines. In particular, the system demonstrated the ability to accurately reproduce graphs and results published in scientific articles.

The implementation of such tools allows laboratories to minimize resource costs for re-mastering methods developed by predecessors. The automation of knowledge transfer through LabAgent reduces the time spent on training new employees and increases the reproducibility of scientific discoveries, which is critical for long-term research projects in the life sciences.