Combining Multi-Modal Mass Spectrometry Measurements and Computational Predictions for Reference-Free Compound Identification

Metabolomics involves the identification and quantification of small molecules in biological systems to understand their role in an organism’s phenotypic expression. Mass spectrometry (MS) based measurements are preferred because of their sensitivity, resolution, and speed. Efforts over the last 20 years have been dedicated to developing standardized techniques, repositories, and tools for data analysis. With these advancements however, there remains a sharp limitation in compound identification. MS does not directly measure molecular structure and as a result relies on matching observed signatures to those of reference compounds for identification. This presents a significant limitation, as the majority of small molecule chemical space is not available as reference standards. Recently, we developed a new methodology that includes advanced instrumentation and computational capabilities for the reference-free identification of metabolites. Our methodology relies on the use of multi-modal measurements including exact mass, tandem mass, collisional cross section, and infrared measurements to precisely describe an ion of interest. Measured properties are then compared to calculated properties of several candidate molecules for a probability-based identification. Here we demonstrate this approach for the reference-free identification of 10 fentanyl analogs within a blinded challenge sample. Specific to this work was the prediction and prioritization of ~1 billion novel potential analogs that were added to the known fentanyl chemical space (~60,000). We find the ability to correctly identify 5 analogs as the top candidate, 3 within the top 2, 1 within the top 5, and 1 within the top 10. These results show 80% of the measured analogs are correctly identified within the top 2 candidates when starting from a candidate list of > 1 billion. This use-case provides line of sight for broad utility of reference-free compound identification in metabolomics and related studies.