AI agents guided by logic rules boost reproducibility in automated lab discovery
Researchers built an AI system that combines language models with logical guardrails to run experiments autonomously, identifying new cellular interactions in yeast. The work suggests that constraining AI with structured knowledge—rather than letting it run freely—improves reliability and could accelerate drug discovery pipelines that currently depend on manual hypothesis-testing.
Originaltitel: Agentic AI integrated with scientific knowledge: laboratory validation in systems biology
**Agentbaserad AI med logisk styrning automatiserar systembiologisk upptäckt** LLM-baserade agenter kopplade till laboratorieinstrument kan validera biologiska hypoteser utan mänsklig växling mellan experiment och analys – förutsatt att man bygger in logisk styrning. Chalmers och KTH presenterar en integreringsmodell där språkmodeller guidats av symbolisk relationslearning och kontrollerade vocabulärer samarbetar med automatiserad cellodling och metabolomik. Systemet identifierade tidigare okända fenomen i jäst: glutamatinducerad tillväxthämning vid sperminbehandling och aminoadipats partiella räddning vid formylsyrestress. Upptäckterna lagrades i grafdatabas med semantisk representation enligt beskrivningslogik. Ramverket löser ett kritiskt FoU-problem — LLM:ers tendens till logiska brister — genom strukturerad ontologiintegration. För biotech-aktörer relevants: systemet minskar cyklustid mellan hypotes och validering och skapar återanvändbara experimentella modeller.
Abstract Automation is transforming scientific discovery by enabling systematic exploration of complex hypotheses. Large language models (LLMs) perform well across diverse tasks and promise to accelerate research, but often struggle with logical structures. Here, we present a framework for biological discovery integrating LLM-based agents with laboratory automation, guided by logical scaffolds incorporating symbolic relational learning, structured vocabularies and experimental constraints. This integration improves coherence and reliability in automated workflows. We couple this AI-driven approach to automated cell-culture and metabolomics platforms, enabling integrated hypothesis validation and refinement, yielding a flexible discovery system. The system identified novel interactions in Saccharomyces cerevisiae, including glutamate-induced growth inhibition in spermine-treated cells and aminoadipate’s partial rescue of formic-acid stress. All hypotheses, experiments and data are captured in a graph database employing controlled vocabularies. Existing ontologies are extended, and a novel representation of scientific hypotheses is presented using description logics. This work demonstrates the potential for a reliable machine-driven discovery process in systems biology.