Longitudinal Fragment Profiles Based on Multi-Collision Energy Tandem Mass Spectra Improve the Accuracy of Metabolite Identification in Untargeted Metabolomics.

Metabolite identification in untargeted metabolomics via tandem mass spectrometry (MS/MS) spectral matching is commonly performed by comparing experimental and reference MS/MS spectra acquired at one or a few collision energies, which generates a similarity score based on the relative intensities of fragment ions within each spectrum, referred to as cross-sectional profiling. Here, we introduced a novel method that significantly improved identification accuracy by comparing longitudinal fragment profiles, which consisted of the intensities of individual MS/MS fragments across multiple collision energies. This approach, termed longitudinal profiling, highlighted low-abundance fragments that were often overlooked by conventional cross-sectional methods, emphasizing predominant ions. We optimized the Jaccard similarity algorithm for longitudinal profiling and established identification criteria using an in-house spectral database comprising approximately 1,80,000 MS/MS spectra. The robustness of the method was validated using inter-instrument datasets, spiked standards, and human plasma samples. Compared with cross-sectional profiling using the optimal entropy algorithm, the longitudinal profiling method improved annotation accuracy by 8.7-25.9% and reduced the false discovery rate by 28.6-41.7%, resulting in a fair increase in the number of confidently annotated metabolites. This method enhances the probability of discovering true diagnostic markers while reducing the likelihood of false diagnostic markers. Our results demonstrate that longitudinal profiling provides a promising new avenue for more accurate metabolite identification in untargeted metabolomics.