Unknown compounds analysis by HRMS, soft ionization and AI for GCxGC (Masaaki Ubukata, MDCW 2025)

🎤 Presenter: Masaaki Ubukata (JEOL, Tokyo, Akishima, Japan)

💾 PDF presentation

💡 Book in your calendar: 17th Multidimensional Chromatography Workshop (MDCW) 13 - 15. January 2026

Abstract

GC-MS plays an important role in a lot of application fields and is widely used for qualitative and quantitative analysis, especially for volatile organic compounds. Most of the qualitative analysis in GC-MS is compound identification by comparison with commercially available library databases for the EI mass spectra. The EI mass spectral pattern, which is commonly used in GC-MS, is known to be highly reproducible and independent of the instrument. Therefore, a large number of EI mass spectra are available in commercial databases, and it becomes an advantage comparing with qualitative analysis in LC-MS.

However, there are still many compounds that are not registered in commercial EI mass spectral databases. For this qualitative analysis issue, we have created an AI model to predict EI mass spectra from structural formulas by machine learning.

We prepared approximately 100 million compounds structures from PubChem, 10 million TMS compounds and 5.5 million pyrolyzates compounds using in-silico. Totally, we have created a new predictive EI mass spectral database for 120 million compounds that are not registered in commercial databases.

In this study, we report the details of the development for the predicted EI mass spectral database and applying to comprehensive 2-dimentional gas chromatography data.