Validating Centralized Biobanking Workflows for NMR Metabolomics Using the PRIMA Panel

Preanalytical factors, such as delays in centrifugation and freezing, can significantly impact metabolomics research by altering metabolite concentrations. This study examined controlled delays in blood samples collected in different tube types, quantifying metabolites using NMR spectroscopy and modeling their stability. The concept of a "stability time point" was introduced to mark when metabolite levels change by 20%. 

The PRIMA Panel, an open-source R Shiny tool, was developed to help researchers assess these effects and generate performance reports. Validated in a biobank setting, the tool enhances quality control and ensures reliable metabolomics analyses.

The original article

Validating Centralized Biobanking Workflows for NMR Metabolomics Using the PRIMA Panel

Heidi Altmann, Marko Barovic, Katrin Straßburger, Maximilian Tschäpel, Sophie Jonas, David M. Poitz, Alexia Belavgeni, Triantafyllos Chavakis, Peter Mirtschink and Alexander M. Funk

Anal. Chem. 2025, 97, 5, 2762–2769

https://doi.org/10.1021/acs.analchem.4c04938

licensed under CC-BY 4.0

In modern centralized biological repositories (biobanks), it is recommended to collect comprehensive preanalytical information for each submitted sample to effectively control for potential confounding factors and errors. This includes time-stamps for sample collection, centrifugation and freezing times. (17) Such preanalytical information can be categorized using the SPREC (Standard PREanalytical Code), which is a seven-letter code that classifies the preanalytical quality of a sample. (18) Each letter represents a different preanalytical condition, including sample material, type of tubes, as well as storage conditions. The pre- and postcentrifugation delays are categorized by letters 2 and 6, respectively. These letters categorize processing delays ranging from <30 min (A1) to >48 h (H) and indicate whether the delays occurred at room temperature (RT) or in a cooled state (2–10 °C). In a clinical setting or a centralized biobank the pre- and postcentrifugation delay can vary substantially from sample to sample. (19−21)

Measurements of metabolic parameters from such a collection can be problematic, due to the potential range of these processing delays. Here, we investigated the detailed effects of prolonged pre- and postcentrifugation times on metabolic parameters using NMR Spectroscopy with the objective of establishing a sample quality control tool for sample cohorts with varying processing delays. First, experiments were performed to observe the direct effect of prolonged pre- and postcentrifugation delays. These data were then used to generate linear mixed models that allow for prediction and quantification of the changes each metabolic parameter undergoes with prolonged processing delays. Finally, we aimed to validate samples obtained from a preanalytical workflow that contain variable pre- and postcentrifugation delays (e.g., centralized biobank) for NMR-based metabolomics. Here, we introduce the PRIMA Panel (Pre-Analytical Investigator for NMR-based Metabolomics). The PRIMA Panel, an open-source tool, was created to allow researchers to directly assess the preanalytical information on their sample cohorts and observe the effects of preanalytic variability on metabolic parameters. Bruker BioSpin offers a commercially available tool termed BiobankQC, for plasma/serum and urine samples analyzed using NMR spectroscopy and their IVDr methods. This panel provides an overview of proper sample preparation, potential contamination, and the accuracy of NMR measurements. Additionally, it detects the anticoagulant used in collection tubes (e.g., citrate) and evaluates matrix integrity parameters (e.g., lactate, glucose), which reflect preanalytical variations such as bacterial growth or improper temperature handling during preanalytical procedures. The panel also assesses changes in the concentration and chemical shift of rapidly varying metabolites like glucose and lactate caused by prolonged preanalytical delays. However, it only delivers a basic pass/fail output for these parameters. PRIMA was developed to enhance this functionality, offering a broader range of information on processing delays, and a user-friendly overview tailored for biobank users.

A cohort from the Biobank Dresden (BBD) was used to highlight a realistic range of preanalytical variations and provide an example workflow for the PRIMA Panel. The PRIMA Panel could prove useful in cases where samples were collected in a unique clinical situation with no possibility of repeating sample collection.

Experimental

NMR Spectroscopy

Sample preparation and ¹H NMR spectroscopy measurements were performed according to established protocols (14,16) and following the Bruker IVDr Methods protocols. All NMR measurements were carried out on a Bruker BioSpin 600 MHz AVANCE NEO equipped with a BBI-Probe (5 mm) and a Bruker SampleJet robot with a cooling system for sample storage at 4 °C. A full quantitative calibration was performed before each measurement. The panels used for the metabolic profiles were B.I.Quant-PS (version 2.0.0) and B.I.LISA (version 1.0.0), which include up to 42 metabolic parameters and 112 lipid subclass parameters. The names of the B.I.LISA parameters were changed for clarification (Table S1). The BiobankQC-PS (version 1.0.0) was also measured.

Analysis and Programming

Metabolites with more than 30% missing values were removed from the analysis. Statistical analysis and model creation were performed in R utilizing the packages “rstatix”, “lme4” and “performance”. Models were created with about 2/3 of the measured samples (n = 20), and the remaining samples were used for the validation cohort (n = 10 for precentrifugation, n = 7 for postcentrifugation). The linear mixed models were created with cell count as a random slope and donors as random intercepts (Tables S4 and S5). The validation cohort was used to calculate the percentage mean error for the difference between predicted values from the models and actual values of the validation cohort. The PRIMA Panel application was written in the R environment (22) and utilizes the package shiny (23) to display it as a web application. A web version of the application is permanently hosted on the shinyapps.io platform via https://funkam.shinyapps.io/PRIMA/. The source code, as well as the original data tables, are available on GitHub via https://github.com/funkam/PRIMA and available for local deployment.

Results and Discussion

Pre-analytical Investigator for NMR-Based Metabolomics (PRIMA Panel)

Furthermore, an R Shiny web application was created using the previously presented linear mixed models. This application enables users to assess the effects of preanalytic variation on their data and generate performance reports for their sample collection. The tool is hosted online permanently but can also be run locally using the code deposited in GitHub.

The panel is split into two different parts. First, in the “Data” module, the data of the preanalytical variation experiments, the linear mixed models and the stability time points are presented in interactive table and plots. Here, users can set their own threshold (in percent) for the stability time points and the metabolic parameters are automatically categorized into SPREC and sorted according to their time points as a means to explore the data in more detail. The timeline plots analogous to those in Figures 2 and S4 are also reproduced. This part of the panel is designed as a quick reference guide and as an extension of the data presented here.

The “Performance” module of the PRIMA Panel provides a more granular view of the concentration changes of the measured analytes, as well as an overview of the preanalytical information on a sample collection. Users can enter the data for a single sample and explore the effects using an interactive slider (Figure 4A) or upload a file with preanalytical information on a sample collection. A tool is available to prepare a table in the format required for the reports, which automatically calculates pre- and postcentrifugation delays from date-time stamps. To calculate the postcentrifugation delay accurately, it is also possible to set the centrifugation time in case the time stamps were obtained from the beginning of the centrifugation. The tool also generates performance reports containing the SPREC classification and a variety of visualizations that provide an overview of the preanalytic characteristics of the cohort (Figure 4B–D).

Anal. Chem. 2025, 97, 5, 2762–2769 - Validating Centralized Biobanking Workflows for NMR Metabolomics Using the PRIMA Panel - Figure 4. “Performance” module of the PRIMA Panel. (A) Snapshot of the user interface, and presentation of the analytical plots for the reduced DILB cohort of just serum samples (B–E). Histogram, density plot, and SPREC distribution are produced automatically.

The final part of the PRIMA Panel involves predicting levels of metabolic parameters. Here, in vitro changes in concentrations of metabolic parameters are estimated using a combination of two different linear mixed models for pre- and postcentrifugation delays. The prediction results are presented as a dynamic table that users can customize for colors and thresholds (Figure 4E). The MPE, as mentioned above is included as an interactive input variable and parameters can be filtered using the “Tolerable Error” panel. In summary, users upload a table with datetime-stamps for collection, centrifugation and freezing and the tool creates automated analytical plots, as well as provides information about the metabolic changes expected for each sample. The output can be saved as an interactive HTML document. The PRIMA Panel is permanently hosted online (https://funkam.shinyapps.io/PRIMA/). The source code is openly available and can be deployed locally.

Conclusion

In this study, a combination of pre- and postcentrifugation delays was modeled to obtain a comprehensive overview of preanalytical process for metabolomics samples in the setting of a centralized biobank. The models were used to identify metabolic parameters that change due to these processing delays. A validation cohort was used to investigate the robustness of these models. Next, these models and data were combined to create the PRIMA Panel, which allows the exploration of the data and provides output in form of performance reports. The data presented here highlight the importance of tracking preanalytical information for biobanking repositories. Additionally, it provides a flexible tool to investigate the direct effect of preanalytic delays on metabolic parameters facilitating data driven filtering of samples and parameters to increase the reliability of subsequent analyses.