Overcoming the Limitations of Mass Spectrometry Across Datasets
Metabolon's global assay technology can process thousands of samples while delivering longitudinal reproducibility
“Clinical and epidemiological studies often involve the analysis of hundreds to thousands of samples across studies or as part of the same study at different time points. This process may represent a significant challenge when performing untargeted mass spectrometry (MS)-based metabolomics.” -Giuseppe Astarita, Senior Scientific Manager, Metabolon.
Mass spectrometry is an inherently semi-quantitative, highly sensitive technology that measures the relative quantity differences of an individual metabolite as expressed by the metabolite’s peak intensity variations in comparative samples. However, this relative quantitation presents a limitation in untargeted MS-based metabolomics when drawing comparisons between samples processed on different runs of the same instrument or other runs from different instruments.
At Metabolon, we implemented solutions based on data normalization and quality control (QC) as critical data processing components to improve data quality and reduce the non-biological variation in the samples. These solutions allow us to overcome some of the limitations of MS-based metabolomics while maximizing longitudinal reproducibility.
Approaches to normalization
“In metabolomics, like other ‘omics, data normalization and quality control (QC) represent critical components of data processing to improve data quality and reduce the non-biological variation in the samples, such as instrument run day drift,” says Matt Mitchell, Senior Director of Statistics, Metabolon.
Sample normalization in metabolomics is key to deriving accurate biological insight, but care must be taken because of the diversity of metabolite structures and behaviors. There are different normalization approaches, including methods that adjust:
- Based on the signal intensity of the sample, for example, by dividing the intensity value of each feature or peak detected in the sample by the total intensity of the entire sample (TIC normalization);
- Based on the individual signal intensity of each metabolite. Examples include 1) dividing the intensity value of each metabolite by its median intensity across the experimental samples or 2) dividing the intensity value of each metabolite by its median intensity in control samples. The control samples represent a QC matrix that ideally should be pooled from representative samples of the study population (if this option is not available due to sample limitations or feasibility, Metabolon maintains pooled QC matrices for several different samples types).
Metabolon has been focused on the quality of its metabolomics data from its early days – more than 20 years ago. “Metabolon’s commitment to the highest quality data has attracted industry recognition,” says Annie Evans, Ph.D., Director of Research and Development for Metabolon. Read more about our participation in the Metabolomics Quality Assurance & Quality Control Consortium (mQACC) working to develop and teach industry standards here.
In untargeted metabolomics, there is no standard method for measuring the total amount of metabolites directly, however, Metabolon has performed extensive analyses in this regard supported by publications and found that the abovementioned second approach greatly outperformed other methods (1).
“When performing normalization to metabolomics data, it is important that the method appropriately corrects for the systematic variation but preserves the biological variation,” says Greg Michelotti, Senior Scientific Director of Biology, Metabolon.
In a 2018 study, we determined the best way to normalize metabolomics data based on analysis of plasma samples obtained from participants in the Insulin Resistance Atherosclerosis Study (IRAS) (1). From this cohort, 1,716 samples were analyzed using Metabolon’s global LC-MS metabolomic profiling platform. Accommodating this many samples required between 13 and 15 instrument runs per arm of the platform. The resulting analysis measured 1,274 metabolites. Untargeted metabolomic profiling was compared to a separate targeted assay for a subset of metabolites representative of multiple biochemical classes. In this study, we showed that the normalization methods that rely on metabolite-specific adjustments significantly outperformed the methods that make adjustments across each sample, such as total ion count (TIC) normalization.
Deeper into the normalization process:
In many cases, the sample-based normalizations performed worse than performing no normalization. Correcting by the median batch value from the experimental samples (MED) can work well in various applications, i.e., for each metabolite, divide the raw peak areas for a sample by the median of the raw peak areas for all samples in the same instrument batch.
However, suppose one wants to run a very small set and merge it into previous data sets or compare the values in two different data sets. In that case, it is typically better to normalize by bridging control samples (BRDG): for each metabolite, divide the raw peak areas for a given sample by the median of the raw peak areas of the bridging control samples. The main drawback of BRDG is that metabolites that are not present in the bridge samples cannot be normalized (1). This publication provides an in-depth comparison of various normalization methods for LC/MS metabolomics data
Approaches to QC
To minimize technical variability when analyzing thousands of samples, Metabolon adheres to QC guidelines set by the NIH’s Metabolomics Quality Assurance & Quality Control Consortium (mQACC) (3,4). For example, in a recent study led by Claudia Langenberg at the MRC Epidemiology Unit, the University of Cambridge that has been published in Nature Medicine, measurements were undertaken in two sub-cohorts of 5,989 and 5,977 participants, respectively. In total, 1,015 metabolites were measured in both sub-cohorts (2). Several types of controls were analyzed in concert with the experimental samples: a pool of well-characterized human plasma served as a technical replicate throughout the dataset; extracted water samples served as process blanks; and a cocktail of QC standards that were carefully chosen not to interfere with the measurement of endogenous compounds was spiked into every analyzed sample, allowed instrument performance monitoring and aided chromatographic alignment. Instrument variability was determined by calculating the median relative standard deviation for the standards that were added to each sample before injection into the mass spectrometers. Overall process variability as determined by calculating the median relative standard deviation for all endogenous metabolites (that is, non-instrument standards) present in 100% of the pooled matrix samples was 10%. Experimental samples were randomized across the platform run with QC samples spaced evenly among the injections (2).
At Metabolon, we process thousands of samples, and we can integrate data from different studies and among other patients over time into the same dataset. To do so, we developed solutions to overcome some of the limitations of mass spectrometry.
When performing normalization to metabolomics data, it is important that the method appropriately corrects for the systematic variation but preserves the biological variation. Methods that rely on metabolite-specific correlations perform better than sample-based normalizations.
In addition, “at Metabolon, we are CAP and ISO accredited and adhere to rigorous QA/QC guidelines set forth by these agencies. Many of these guidelines are in place to help monitor and overcome the limitations of MS-based untargeted metabolomics analysis.” -Annie Evans, Senior Director R&D at Metabolon.
Metabolon’s competency to deliver longitudinal reproducibility translates into high-quality data for clients.
Learn how Metabolon can help accelerate your research, contact us today.
- Wulff, Jacob E., and Matthew W. Mitchell. “A comparison of various normalization methods for LC/MS metabolomics data.” Advances in Bioscience and Biotechnology 9.08 (2018): 339.
- Pietzner, Maik, et al. “Plasma metabolites to profile pathways in noncommunicable disease multimorbidity.” Nature medicine 27.3 (2021): 471-479.
- Beger RD, Dunn WB, Bandukwala A, Bethan B, Broadhurst D, Clish CB, Dasari S, Derr L, Evans A, Fischer S, et al. (2019). Towards quality assurance and quality control in untargeted metabolomics studies.Metabolomics.15, 4–.
- Evans AM, O’Donovan C, Playdon M, Beecher C, Beger RD, Bowden JA, Broadhurst D, Clish CB, Dasari S, Dunn WB, et al. (2020). Dissemination and analysis of the quality assurance (QA) and quality control (QC) practices of LC-MS-based untargeted metabolomics practitioners.Metabolomics.16, 113–.