Evaluation of Signal Response for Data Interpretability in UHPLC–HRMS and Other Applications

News
Event

Three different metrics were used experimentally, based on measurements of both NIST Standard Reference Material (SRM) 1950 and data obtained from murine kidney tissue.

A study recently published in the Journal of the American Society for Mass Spectrometry presents a strategy for the signal response evaluation of untargeted mass spectrometry (MS) data—in particular, according to the authors, ultrahigh-pressure liquid chromatography (UHPLC) coupled to high-resolution mass spectrometry (HRMS) (1). The objective the researchers aimed to achieve was to remedy shortcomings in feature finding, or feature filtering, meant to identify differences in sample groups. Too often, they said, the minimization of false discoveries in feature filtering leads to the filtering out of true discoveries.

The doctor looks at the kidney hologram, checks the test result on the virtual interface and analyzes the data. Kidney disease, stones, innovative technologies, medicine of the future. | Image Credit: © Aliaksandr Marko - stock.adobe.com

The doctor looks at the kidney hologram, checks the test result on the virtual interface and analyzes the data. Kidney disease, stones, innovative technologies, medicine of the future. | Image Credit: © Aliaksandr Marko - stock.adobe.com

The authors, Kirsten E. Overdahl, Justin B. Collier, Anton M. Jetten, and Alan K. Jarmusch, all of whom are affiliated with the Immunity, Inflammation, and Disease Laboratory in the National Institute of Environmental Health Sciences at the National Institutes of Health in Durham, North Carolina, USA, demonstrated their hypothesis in metabolomics measurements from two sources, one a uniform biological matrix, namely the National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1950, human plasma, and the other source being a variable biological matrix (murine kidney tissue) (1).

As the authors define in their study, untargeted MS encompasses courses of study such as metabolomics and lipidomics, but the myriad ways in which the technique is employed make data processing and interpretation quite complex (1). Feature-finding software, they said, can be of assistance in separating data into key descriptors such as mass-to-charge ratio (m/z), retention time, and mobility.

A portion of this study was spent reviewing prior attempts in the lab and literature to solve the dilemma of feature measurement. To establish this conundrum, the authors explained that a relationship that is fundamental to all MS measurements is that which exists between concentration and sample, which is both matrix- and analyte-specific and influenced by electrospray ionization (EI) (1). In this evaluation of previous work, the researchers mentioned several attempts at a quality control (QC) dilution series, but said that even with individual feature response profiles, linear response assessment tended to skew Pearson correlation values, emphasizing the need for additional solutions.

To that end, the research team supplemented Pearson correlations, which evaluate linear response, with two other metrics including Spearman’s ρ, for monotonic response, and coefficient of determination R2, to fit to a linear model (1). The reason for the additional metrics was the hypothesis that if more material was present, the signal would correspondingly increase, providing flexibility in data assessment.

A summary of the team’s findings highlighted one result in particular from this approach, which was that a chemical the researchers deemed “putatively exogenous,” ethylenediaminetetraacetic acid (EDTA), in the case of NIST SRM 1950, was removed from downstream consideration, supporting the theory that features not “true” of the biological sample would be taken out in these processes. With regard to the murine kidney tissue, filtering resulted in a stronger correlation between intergroup separation and total observed variance. The researchers recommended their method of evaluation in untargeted MS and, because a priori chemical knowledge and standards are not required, suggested it could be used broadly in the future.

Reference

(1) Overdahl, K. E.; Collier, J. B.; Jetten, A. M.; Jarmusch, A. K. Signal Response Evaluation Applied to Untargeted Mass Spectrometry Data to Improve Data Interpretability. J. Am. Soc. Mass Spectrom. 2023, 34 (9), 1941–1948. DOI: 10.1021/jasms.3c00220

Recent Videos
Toby Astill | Image Credit: © Thermo Fisher Scientific
John McLean | Image Credit: © Aaron Acevedo
Related Content