Symposium Sessions
Day 2, June 11（Thu.） 9:55-10:15 Room B (4F 411+412)
- 2B-S1-0955
Hivebot: An Interactive AI Assistant for Mass Spectrometry Data Analysis
(Reifycs)
Ipputa Tada, Kazuto Mannen, oMitsuhiro Kanazawa, Atsushi Ogiwara
Large language models (LLMs) and other AI technologies are increasingly applied to research support and data analysis. However, a general-purpose AI system capable of directly handling mass spectrometry (MS) data and executing analytical operations through interactive dialogue has not yet been widely established. MS measurement data are typically stored in manufacturer-specific proprietary formats, and conversion into XML-based formats often increases data size and reduces processing speed. In addition, advances in MS instrumentation have led to rapid growth in data volume, making data conversion and access a computational bottleneck for AI-driven analysis.
In this study, we developed Hivebot, an interactive AI assistant for MS data analysis. By utilizing a manufacturer-independent binary data access platform (Hive), efficient access to MS data was achieved through a unified API. Hivebot enables users to visualize spectra and chromatograms, perform peak detection, and conduct multivariate analyses through natural language instructions. Furthermore, a machine learning–based peak detection model trained on manually identified peaks was integrated into the assistant. The results demonstrate that interactive AI-assisted MS data analysis can be implemented with practical analytical accuracy.