MetaboT: An LLM-based Multi-Agent Framework for Interactive Analysis of Mass Spectrometry Metabolomics Knowledge 1. MetaboT introduces a novel multi-agent framework that leverages Large Language Models (LLMs) to convert natural language questions into precise SPARQL queries, enabling intuitive and efficient navigation of complex metabolomics knowledge graphs. This innovation significantly lowers the barrier for researchers without extensive programming skills to explore and analyze mass spectrometry metabolomics data. 2. The framework stands out for its modular design, which allows for straightforward extension to other mass spectrometry-based knowledge graphs and compatibility with different LLM models. This flexibility ensures that MetaboT can adapt to the rapidly evolving landscape of LLMs and integrate new capabilities seamlessly. 3. MetaboT's performance was rigorously validated on a large plant dataset with 50 representative queries. The results demonstrated a substantial improvement in SPARQL query accuracy when using the multi-agent system compared to single-LLM baselines, highlighting the effectiveness of the coordinated functioning of specialized agents in mitigating LLM limitations such as hallucinations. 4. A notable feature of MetaboT is its iterative refinement loop, which distinguishes between query construction errors and genuine data absence. This mechanism reduces manual debugging efforts and enhances user confidence by automatically reformulating and re-executing queries when necessary. 5. The system's architecture includes a flexible pipeline that manages entity resolution, query processing, and iterative refinement. Specialized agents and tools are employed for tasks such as verifying the presence of plant species in curated databases, resolving entity identifiers through authoritative sources, and generating executable SPARQL queries aligned with the knowledge graph's schema. 6. MetaboT's web application interface prioritizes usability, capturing natural language questions and displaying results while managing the backend processing. Current features include file uploads, visualizations, and a SpectrumPlotter module for integrating Metabolomics USIs to render spectrum plots. 7. Future developments aim to address limitations such as the current restriction to single-graph querying and the computational cost associated with multiple LLM API calls. The vision is to evolve MetaboT into a comprehensive toolbox for mass spectrometry data analysis by integrating agents for common computational metabolomics approaches and expanding interoperability with other tools. 📜Paper: arxiv.org/abs/2510.01724 💻Code: github.com/HolobiomicsLab/Me… #MetaboT #LLM #MultiAgentSystem #MassSpectrometry #Metabolomics #KnowledgeGraphs #SemanticDataIntegration #ComputationalBiology

WANTED at @UM_IDS Postdoctoral researcher in Biomedical Discovery Science. academictransfer.com/en/3225… @micheldumontier #umdatascience #datascience #ai #personalisedmedicine #datacuration #semanticdataintegration #naturallanguageprocessing #ontologyengineering #predictivemodelling

Great news! @CPSoSAWARE joins the @ontocommons demonstrators with its #ontology-based #SemanticDataIntegration solution for the #manufacturing use case! More news coming soon!

#semanticDataIntegration makes validation easier. Eg in @ENTSO_E #electricity data: rightmost red link (from EIC) is wrong, the blue link (from PROD_UNITS) is right re

CASPAR now deployed in an #eHealth @HorizonEU project! A truly domain-agnostic #SemanticDataIntegration framework!

@CASPAR_project went to #Barcelona! The novel #SemanticDataIntegration framework was publicly presented for the first time at #SEMAPRO21. Stay tuned for more news coming soon! #SemanticWeb #KnowledgeGraphs #DataFusion #AutonomousVehicles #DriverMonitoringSystem

Do your #PhD in #ComputerScience #Mathematics or #InformationScience and apply for a positionin in the #ScienfiticDataManagement Group at @TIBHannover ow.ly/xRKN50xv321 #SemanticDataIntegration #MachineLearning

Pursuing a #PhD in #ComputerScience #Mathematics or #InformationScience? @TIBHannover offeres several positions in projects of the #ScienfiticDataManagement Group! ow.ly/8Twi50xv30b #SemanticDataIntegration #MachineLearning

A great example in the fields of #cancerresearch & #pharmacogenomics of applying #semantictechnology, #semanticdataintegration, enriched w/ wider knowledge bases such as #wikidata and powered by #graphdatabase such as @OntotextGraphDB. And there's a demo! github.com/lasircc/semalytic…