Automated compound identification using product ion scanning with accurate mass measurement and compound database searching for non-targeted metabolomics

Importance of the Topic

Non-targeted metabolomics allows comprehensive profiling of small molecules in complex samples but relies on robust compound identification to translate spectral data into meaningful chemical structures. Automated and accurate identification is critical for high-throughput studies in biological, food, environmental and industrial contexts.

Objectives and Study Overview

This work presents an integrated workflow combining high-resolution MSn product ion scanning, accurate mass formula prediction, and database searching to automate candidate ranking in non-targeted metabolomics. The approach was evaluated on green tea leaf extracts to identify metabolites that drive tea quality, using multivariate modeling to link chemical profiles with sensory rankings.

Methods and Instrumentation

Green tea leaves were extracted, filtered and analyzed by liquid chromatography coupled to an ion-trap time-of-flight mass spectrometer (LCMS-IT-TOF) with ESI source operating in polarity switching mode. MS1–MS3 spectra were acquired over m/z 100–1000 and passed to Formula Predictor software for elemental composition scoring. Predicted formulas were used to query an in-house compound database via a ChemSpider™ API interface. Candidate structures were automatically evaluated by matching theoretical and observed product ion spectra.

Used Instrumentation

• Shimadzu LCMS-IT-TOF with ESI (+/–) scan range m/z 100–1000
• Shim-pack XR-ODS column (2.0×50 mm, 2.2 µm) at 40 °C, 0.4 mL/min gradient (0.1% formic acid in water/methanol)
• Formula Predictor (Shimadzu) and custom database search via ChemSpider API

Results and Discussion

From ~3 742 detected features, isotopic and statistical filtering yielded 462 peaks for modeling. A PLS regression model correlated spectral features with tea sensory scores (R2≈0.97). The top 20 variables by VIP were subjected to automated identification. Example feature var_337 generated 218 formula candidates, which were narrowed to 6 top hits by combined mass accuracy and product ion assignment scoring. Identified compounds included epicatechin gallate isomers and related phenolic esters central to tea quality.

Benefits and Practical Applications of the Method

The automated workflow eliminates laborious manual spectral interpretation and literature searches, streamlining metabolite identification in complex matrices. It supports rapid QA/QC and biomarker discovery in food, pharmaceutical and environmental analyses.

Future Trends and Potential Applications

Enhancements may include integration of machine learning for improved scoring, expansion of compound libraries to cover additives, contaminants and polymer residuals, and extension to higher-throughput platforms. Broader adoption could accelerate non-targeted screening in regulatory compliance, natural product research and clinical metabolomics.

Conclusion

The combined use of accurate mass MSn acquisition, formula prediction and automated product ion matching provides a powerful, generalizable strategy for non-targeted metabolite identification. Applied to green tea quality assessment, it successfully prioritized key phenolics and demonstrates potential across diverse analytical fields.