Real-Time Quality Control Software for Metabolic Phenotyping
App Note / Case Study
Published: October 25, 2023
Metabolomic profiling via mass spectrometry generates complex multidimensional data from thousands of potential metabolites and large sample cohorts, making data quality control (QC) an essential yet demanding process.
QC can be performed after data acquisition, but unexpected deviations in data quality can waste valuable time, samples, and resources. Instead, RealTime-QC allows for active monitoring of instrument performance metrics as the analysis is being performed, preventing errors that can compromise the analysis.
This application note explores how the latest QC software supports real-time analysis for metabolic phenotyping.
Download this application note to learn about:
The challenges associated with metabolomic data collection
The value of Real Time-QC software
Insightful instrument performance metrics that can help you improve your data quality
Real-time and post-acquisition quality control (QC) in metabolic phenotyping studies Aiko Barsch, Patrick Groos, Nikolas Kessler, Matthias Szesny, Sven W. Meyer, Ilmari Krebs, Heiko Neuweger, Matthew R. Lewis; Bruker Daltonics GmbH & Co. KG, Bremen, Germany. Abstract • Metabolic profiling of biofluids using hyphenated mass spectrometry-based techniques including LC-MS and LC-TIMS-MS is a workflow of prime importance for biomarker research. • High precision measurements are a prerequisite for deriving meaningful conclusions from multidimensional data. • Both random and systematic errors are possible in each dimension of measurement, creating a need for thorough quality control (QC) monitoring and post-acquisition data correction of complete feature-sets. • The RealTimeQC module of TASQ® 2023b software runs in parallel with data acquisition, providing detailed insight to instrument performance and data quality based on user-defined QC analytes. • Following data acquisition, MetaboScape® enables correction of feature-specific run order effects and facilitates in-depth analysis of individual feature quality. Keywords: Metabolomics, Quality control, Real-time QC, MetaboScape, TASQ Automatic data processing for QC review Increased data quality by feature specific run order intensity correction Intuitive investigation of data accuracy and precision Intensity RSD based feature filtering for quality focus Readily reveals outliers and trends over time for stop/go decisions Fully integrated interactive assessment of feature qualities RealTimeQC TASQ® Benefits of High quality monitored raw data MetaboScape® Benefits of During data acquisition Post-data acquisitionIntroduction Modern methods for the global profiling of metabolites in complex biological samples utilize powerful bioanalytical systems that combine high resolution mass spectrometry (HRMS) with ultra-high performance liquid chromatography (UHPLC) and/or trapped ion mobility spectrometry (TIMS) for highly selective metabolite measurement. Together, these systems produce complex multidimensional data for hundreds to thousands of metabolites across similar numbers of samples, with both random and systematic errors possible in each dimension of measurement. The resulting data is complex, and consequently, evaluation of the data quality is challenging. To make matters worse, as outlined by Broadhurst et al. in 2018 , compared to targeted assays there are no QC acceptance criteria for all the detected metabolites in untargeted metabolomics. For these workflows, the repeated analysis of pooled QC samples can provide relative measures of precision and be used to correct for systematic measurement error including run order-based drift in signal intensity. However, such analysis is conventionally performed after data acquisition is completed, by which time any substantial measurement drifts or sudden deviations may be beyond remedy, potentially compromising the analysis. For these reasons, researchers may wish to more closely monitor the quality of their profiling analyses in real-time (i.e. in parallel with data acquisition). However, automated tools to support meaningful real-time investigation of complex multidimensional data are severely lacking in the field. To help tackle these challenges, we introduce here the RealTimeQC module of TASQ 2023b software for in-depth analysis of data quality during acquisition. We further demonstrate Bruker's complete QC workflow which includes MetaboScape for post-acquisition review and correction of profiling data. Methods Six vials were prepared with a 1:5 dilution of standard reference material urine (NIST SRM 3672) in ultra-pure water and 50 injections were performed per vial. This resulted in a total of 300 injections of the diluted sample acquired with an Agilent 1290 Infinity II UHPLC and Bruker impact II VIP QTOF high-resolution mass spectrometer. HRMS acquisition was performed in positive ionization mode featuring a VIP-HESI source (details see Table 1). A 15-minute reversedphase UHPLC gradient was adapted from the National Phenome Center open LC-MS platform [2, 3] (Details see Table 2). Every tenth analysis was labeled (and is herein referred to) as a QC sample, mimicking the conventional QC format of global profiling studies. Data quality was monitored in parallel with acquisition by inspection of endogenous metabolites chosen to represent the broader metabolome (“QC analytes”) using the RealTimeQC module of TASQ 2023b software. Following data acquisition, MetaboScape 2023b was used to perform post-acquisition retention time alignment, mass recalibration, and compensation for run order effects in peak intensity measurements using a soft LOESS correction curve based on feature-specific intensities in designated QC samples. Table 1 MS acquisition parameters MS impact II VIP Source VIP-HESI source End Plate Offset 500 V Capillary 4500 V Nebulizer 2.0 Bar Dry Gas 8.0 L/min Dry Temp 230°C Probe Gas Temp 400°C Probe Gas Flow 4.0 L/min Ionization Positive ion mode Acquisition mode MS – 5 Hz Transfer parameters Scan range 20-1300 m/z Funnel 1 RF 200 Vpp Funnel 2 RF 200 Vpp Hexapole 50 Vpp Quadrupole Low Mass 60 m/z Quadrupole Ion Energy 5 eV Collision RF 450 V Transfer Time 80 µs Pre Pulse Storage 5 µs Calibration Automatic internal mass calibration using sodium formateRealTimeQC enables real-time monitoring of multidimensional data QC monitoring can be improved using RealTimeQC by first selecting QC analytes endogenously present in or synthetically added (i.e. internal standards) to samples that act as representative markers for the broader metabolic profile and then monitoring them in depth across the analysis. Here we investigated eight such QC analytes selected to cover a representative range of masses and retention times. For each newly analysed sample RealTimeQC provides a customizable array of data quality metrics for each QC analyte (see Figures 2 and 3). Here we investigated the peak area, peak width, mass accuracy, retention time accuracy, and isotopic fidelity across all samples. Table 3 highlights the high raw data quality for the eight investigated QC analytes among the 300 total injections. For all QC analytes, the RSD of peak areas is below 10%. The standard deviations for retention time and mass error are below 0.25 seconds and 1 ppm, respectively. 12 Figure 1 Overlay of Total Ion Chromatograms of every tenth datafile from 300 consecutive analysis of a human urine sample. Conventional QC is performed by manually overlaying chromatograms from the repeated analyses of QC samples. This is often laborious, time consuming, and ultimately lacks information as investigation is typically biased towards only the most intense peaks and lacking the depth and dimensionality produced by modern hyphenated techniques. x107 1.5 1.0 0.5 2 4 8 6 10 Time [minutes] Intensity Conventional QC by overlaying chromatograms is biased towards only the most intense peaks. Results and discussion Conventional QC is tedious and ultimately not informative Interspersed acquisitions of pooled QC samples have become a routine practice in laboratories conducting metabolic profiling experiments. Figure 1 illustrates how QC sample data are typically reviewed during acquisition by overlaying the Total Ion Chromatograms (TIC) of all QC samples in the 300 consecutive human urine analyses. Despite being a common approach for data quality assessment, this form of inspection can be laborious and time consuming. Furthermore, this approach is not deeply informative as investigation is typically biased towards gross chromatographic and signal intensity-based trends and fails to detect possible mass shifts, differences in isotopic fidelity or changes in peak areas of characteristic marker metabolites (QC analytes).Additionally, the median score of the mSigma value, which expresses the isotopic fidelity, was very low with a mSigma value <10 of a maximum possible value of 1000, with lower values indicating better isotopic fit. Despite the overall high quality of the raw data, monitoring in real-time was essential to assessing the nature of potential deviations. RealTimeQC provides intuitive visualization for inspection of random and systematic variation revealing possible trends in data over time and highlighting measurement outliers. For example, the left part of Figure 2 shows mass error (delta m/z) and isotopic fidelity (mSigma) plotted for the selected QC analyte isoleucine across the 300 datafiles acquired showing no cause for concern in the ongoing analysis. In Figure 3A the peak area, peak intensity, delta of retention time (RT) and LC peak Full Width at Half Maximum (FWHM) are plotted for the selected QC analyte, isoleucine. The right part of both figures illustrates the overall variation using violin plots and summary statistics. Figure 2 TASQ RealTimeQC visualization – isoleucine TASQ RealTime QC visualization of mass accuracy and isotopic fidelity for isoleucine in 300 consecutive injections. Both plots reveal stable mass accuracy and isotopic fidelity acquisitions for Isoleucine over 300 injections. TASQ RealTimeQC Δm/z [ppm] mSigma Peak Area [% RSD] StdDev Retention time error [sec] StdDev mass error [ppm] mSigma (median score of max 1000) Butyryl-L-carnitine 7.63 0.063 0.483 5.12 Glutamine 3.38 0.01602 0.625 3.04 Isoleucine 3.50 0.240 0.470 1.17 Isovalerylcarnitine 9.86 0.0672 0.486 2.75 Leucine 2.96 0.0768 0.496 1.23 O-Desmethyl- cis-tramadol 7.69 0.0888 0.490 3.05 Phenylalanine 4.39 0.069 0.442 2.90 Uric acid 3.67 0.0507 0.440 8.33 Table 3 Overview of QC results from 300 consecutive injections of a human urine sample. Following automatic peak detection and integration in TASQ, the Statistics Table provides a highlevel overview on data quality for selected QC analytes (e.g. internal standards or endogenous metabolites chosen to represent the broader metabolome).However, at injection 130, RealTimeQC revealed an anomaly in the peak height (intensity), retention time and peak width of isoleucine during the acquisition sequence (Figure 3 A), triggering inspection of isoleucine in TASQ (Figure 3 B). Analysis revealed that while peak width and retention time had changed, the peak area remained constant. Based on this observation the analysis was allowed to proceed, resulting in a subsequent return to normal and excellent overall data quality. Feature-specific signal correction by MetaboScape Real-time monitoring of data quality during acquisition ultimately led to successful completion of the 300-sample batch and helped demonstrate the excellent precision of the Agilent 1290 Infinity II UHPLC and Bruker impact II VIP QTOF mass spectrometer combination. To investigate how representative of the global profiling data our selected QC analytes were, a complete Figure 3 TASQ RealTimeQC allows rapid assessment of QC analyte accuracy and precision across all samples. A TASQ RealTimeQC visualization of peak area, peak intensity, difference in retention time, and peak width (FWHM) vs. run order for the QC analyte isoleucine. The scatter and violin plots (top) clearly illustrate stable peak area across three days measurement time with 3.4% and 3.5% relative standard deviation (RSD) across QC and all other samples, respectively. However, the peak intensity (second from top) plots show a change in peak height for several sequential injections in the sequence. Similar differences were also observed in retention time and peak width (FWHM). B Analysis of the chromatographic data in TASQ supported the data reported by RealTimeQC, clearly showing a changing peak shape that ultimately had no observable effect on the peak area. TASQ RealTimeQC Area Injection 1 Injection 300 ΔRT [min] Intensity Intensity Intensity FWHM [s] A B Run orderFigure 4 Relative Standard Deviation distribution for feature peak areas extracted by T-ReX® in MetaboScape. The % RSD of feature peak areas determined by MetaboScape across all 300 injections. No within-batch correction was applied. The bin count is plotted vs. RSD [%]. The last bin contains all values >80%. 150 200 250 100 350 76.8% >80% 51.9% 50 300 0 0 10 30 60 20 50 80 40 70 RSD bins [%] Median % RSD = 9.56 Bin count post-acquisition analysis of the data was performed using Bruker MetaboScape software, the results of which are shown in Figure 4. The median % RSD of features peak areas determined by MetaboScape across all 300 injections was 9.55%. In total 2448 features were extracted, with 51.9% of all features having an RSD below 10% and 76.8% of all features having an RSD below 20% highlighting the raw data quality without in-batch correction, and the suitability of RealTimeQC and the analyte QC approach to accurately represent a global profiling dataset. Finally, even with high precision measurement, feature specific run order effects can and do occur. Figure 5 highlights such a feature that was automatically within-batch corrected in MetaboScape using a soft LOESS correction curve based on feature areas in QC samples. This selected example is a reminder that even with high quality monitored raw data, systematic variation correction can still positively impact overall data quality in global profiling applications. Figure 5 Within-batch correction in MetaboScape. The Batch Correction (BC) Review plot highlights a feature showing systematic variation across the batch, and the successful application of feature-specific signal correction.Table 2 LC parameters LC Agilent 1290 Infinity II UHPLC Column Avantor ACE Excel 2 AQ (150 x 2.1 mm, 2.0 µm) Column Oven Temp. 45°C Mobile phase A: Water + 0.1% Formic acid B: Acetonitrile + 0.1% Formic acid Pump Seal wash Isopropanol / Water 10% / 90% Gradient Time [min] Flow [mL/min] %B 0.0 0.60 1 0.1 0.60 1 10.0 0.60 55 10.15 0.61 65 10.30 0.63 75 10.45 0.67 85 10.60 0.75 95 10.70 0.8 100 11.00 1.00 100 11.55 1.00 100 11.65 1.00 1 11.70 0.90 1 11.80 0.80 1 11.90 0.70 1 12.00 0.65 1 12.10 0.61 1 12.15 0.60 1 12.65 0.60 1 Multisampler Temperature 4°C Injection volume 2 µl Use vial/well bottom sensing Yes Draw speed 100 µL / min Eject speed 400 µL / min Wait time after draw 1.2 sec Wash solvent Isopropanol / Water 50% / 50% Standard Wash Flush port, 3 se
Brought to you by
Download this App Note for FREE Below
Information you provide will be shared with the sponsors for this content.
Technology Networks or its sponsors may contact you to offer you content or products based on your interest in this topic. You may opt-out at any time.