An IROA Workflow for correction and normalization of ion suppression in mass spectrometry-based metabolomic profiling data

Abstract Ion suppression is a major problem in mass spectrometry (MS)-based metabolomics; it can dramatically decrease measurement accuracy, precision, and signal-to-noise sensitivity. Here we report a new method, the IROA TruQuant Workflow, that uses a stable isotope-labeled internal standard (IROA-IS) plus novel companion algorithms to 1) measure and correct for ion suppression, and 2) perform Dual MSTUS normalization of MS metabolomic data. We have evaluated the method across ion chromatography (IC), hydrophilic interaction liquid chromatography (HILIC), and reverse phase liquid chromatography (RPLC)-MS systems in both positive and negative ionization modes, with clean and unclean ion sources, and across different biological matrices. Across the broad range of conditions tested, all detected metabolites exhibited ion suppression ranging from 1% to 90+% and coefficient of variations ranging from 1% to 20%, but the Workflow and companion algorithms were highly effective at nulling out that suppression and error. Overall, the Workflow corrects ion suppression across diverse analytical conditions and produces robust normalization of non-targeted metabolomic data.

Metabolite levels reflect metabolic function and the integrated output of genomics, epigenomics, transcriptomics, and proteomics, including inputs from lifestyle and environment [1][2][3] .Hence, metabolomics is an effective approach for elucidating candidate drug targets 4 , candidate biomarkers of disease progression 5 , candidate biomarkers of therapeutic response 6,7 , mechanism(s) of drug sensitivity 8 , mechanism(s) of drug resistance 9 , and mechanisms of drug toxicity 10 .Unfortunately, rigorous, reproducible quantitation of metabolites is difficult.Standardization across laboratories, biological matrices, and analytical conditions is a major challenge for both research and clinical implementation of metabolomics.
Ion suppression is a type of matrix effect in mass spectrometry (MS) and a major contributor to those challenges (Fig. 1).The authors of a recent perspective article on best practices in metabolomics noted: "While there is no universal solution to the ion suppression problem, assessing the effects of ion suppression affords greater confidence in the accuracy of the results." 11Indeed, until now, no universal solution has existed to counteract the negative effects of ion suppression across all analytes in a non-targeted metabolite profiling study.We present such a solution.
Ion suppression for small numbers of analytes (e.g., in targeted MS) can be addressed to some degree by diluting samples, modifying chromatographic or MS conditions to eliminate interferences, conducting a sample cleanup procedure such as solid phase extraction, and/or adding a chemically matched stable isotope-labeled internal standard 11,21 .However, because the source and magnitude of ion suppression can vary extensively across metabolites and samples 11,22 , counteracting ion suppression across all analytes and all samples in a non-targeted profiling study remains an unsolved challenge 11,23 .
Stable isotope-labeled internal standards can correct for variability in ionization efficiency and ion suppression.However, isobaric isotopologs (e.g., the M+0 isotopolog of lactate and the M+1 isotopolog of alanine) are difficult to distinguish.That has been a barrier to the effective use of stable isotope mixtures.Isotopic Ratio Outlier Analysis (IROA) protocols [24][25][26][27] solve that problem by generating clearly identifiable isotopolog patterns.IROA also facilitates removal of non-biological signals, which are common artifacts in MS data.
We introduce here a novel Workflow that effectively corrects ion suppression and uses a Dual-MSTUS (MS Total Useful Signal) normalization algorithm to improve the quantitative accuracy, precision, and signal-tonoise sensitivity of metabolomic data across diverse origins and analytical conditions.

Conceptual
Overview.The Workflow presented here is based on the IROA Internal Standard (IROA-IS) and a chemically identical but isotopically different Long-Term Reference Standard (IROA-LTRS) (Fig. 2).The Workflow identifies each molecule in any type of sample based on a unique, formula-specific isotopolog ladder (Fig. 2C´, D´) created by (i) a low 13 C (natural abundance or 5%) signal from the isotopologs of the molecule in question at the low mass end of the ladder and (ii) a 95% 13 C signal for the isotopologs at the high mass end of the ladder.It is useful to call the collection of lower mass peaks in the isotopolog ladder the 12C channel and the collection of higher mass peaks in the ladder the 13C channel.IROA-LTRS (Fig. 2D, red/yellow samples) is a 1:1 mixture of chemically equivalent IROA-IS standards at 95% 13 C and 5% 13 C.The combination produces the IROA-LTRS isotopic pattern illustrated in Fig. 2D´.Metabolite 12 C and 13 C isotopologs co-elute, and the signature IROA peak pattern distinguishes real metabolites from artifacts, which lack the IROA pattern.
Since metabolites in the Internal Standard are spiked into samples at constant concentrations, the loss of 13 C signals due to ion suppression in each sample can be determined and used to correct for the loss of corresponding 12 C signals.
To model ion suppression, we created a single methanol extract of plasma, divided it into aliquots ranging from 50 to 1,500 µL, dried the aliquots, and reconstituted them with a fixed volume and concentration of IROA-IS.We developed Equation 1 to calculate and correct % ion suppression as described in Methods.
Biologically relevant signals are considered to be those observed in both the IROA-LTRS and analytical samples as an IROA signature isotopolog ladder of signals with regular M+1 spacing, decreasing amplitude signal in the 12 C channel, and increasing amplitude signal in the 13 C channel (from low to high mass) (Fig. 3A).An initial goal of the Workflow for quantitative metabolomic profiling is to calculate an AUC-12Csuppression-corrected value for each metabolite (Equation 1).Without ion suppression, endogenous ( 12 C) metabolite AUCs should increase linearly with aliquot volume, and internal standard ( 13 C) levels should remain constant (Fig. 3B).However, the MSTUS values 28 12 C values increased proportionally to sample input (Fig. 3B).To explore the generalizability of the IROA Workflow, we performed a detailed characterization of ion suppression in both positive and negative ion modes for three highly used chromatographic systems: IC-MS, RPLC-MS (C18-MS), and HILIC-MS with unclean and clean electrospray ionization (ESI) source conditions (Fig. 4A).Negative ionization mode detected fewer ions than did positive ionization mode (Fig. 4B), but extensive ion suppression was evident in both polarities and all chromatographic systems.Ion suppression calculation and corrections were performed automatically by ClusterFinder software (version 4.2.21,64-bit, IROA Technologies) using Equation 1.The list of metabolites present in all samples in each chromatographic system (the Dual MSTUS Metabolite list) is provided in Supplementary Table 1-5.As expected, uncleaned ionization sources demonstrated significantly greater levels of ion suppression than cleaned ionization sources (Fig. 4C-K).
Strikingly, all tested chromatographic systems and conditions exhibited up to nearly 100% ion suppression, which was effectively corrected by the IROA Workflow, which yielded a linear increase in signal with increasing sample input (Fig. 4L, Supplementary Table 6, and Supplementary Fig. 1, 2).Of course, if a metabolite peak were 100% suppressed in the 12 C or 13 C channel, it would not yield the IROA ladder pattern and, therefore, would appear fully suppressed and absent from the output.But even if we attempted to include such a metabolite, the additional unknown in Equation 1 would prevent its correction.Hence, the greatest weakness of the Workflow is that the output includes only metabolites that are detected in both the 12 C and 13 C channels (i.e., at most 99% suppressed).Absence from the 12 C or 13 C channel could be caused by 100% suppression, true absence in the IROA-IS, or true absence from the sample.Future versions of the ClusterFinder software will facilitate identification of fully suppressed analytes.
Historically there has been a tradeoff between injecting a large sample volume to increase assay sensitivity and injecting a smaller volume to decrease matrix effects.Using this Workflow, analysts can err on the side of injecting larger sample volumes to ensure robust measurement of low-abundance analytes while simultaneously performing ion suppression correction to achieve more accurate results.The Workflow produces accurate concentration values for most analytes, even in highly concentrated samples.For example, tryptophan (M+H -NH3) exhibited 8% ion suppression in RPLC positive mode with a cleaned ionization source, and suppression correction restored the expected linear increase in signal with increasing sample input (Fig. 4M).In a more extreme example, pyroglutamic acid exhibited up to 92% suppression in IC-MS negative mode, and the IROA workflow corrected for that suppression as well (Fig. 4N).
The Workflow facilitated identification and measurement of 539 different metabolites across the entire sample set (Fig. 4O).An average of 422 metabolites were observed in each sample, and 216 were common to all samples.Unexpectedly, the total number of peaks observed decreased as a function of sample input, likely due to the suppression driving some compounds' peaks to extinction.Kohonen Self Organizing Maps (SOMs) revealed patterns of increasing concentration with sample input, decreasing concentration with sample input, and various intermediate patterns (Figs.4P, Q and Supplementary Fig. 3).For instance, Figs.4P-15, 4P-13, and 4P-09 indicate metabolite peaks that first increase and then decrease in amplitude as the input volume is increased.Analysis of the full dataset shows: i) for suppression levels up to 60%, the observed [AUC-12Csuppressed] value was weakly correlated with the expected [AUC-12Csuppressioncorrected] value; from 60% to 80% suppression, the correlation was lost; and above 80% suppression, [AUC-12Csuppressed] values were negatively correlated with the expected [AUC-12Csuppression-corrected] value due to decreasing peak size with increasing sample input.
We next evaluated the performance of IROAbased ion suppression correction with urine samples, which notoriously exhibit high ion suppression 5,16 .We were pleased to find that the IROA Workflow produced comparable suppression-corrected data in both positive and negative ionization modes (Figs.4R, S).The correction was robust even at high urine sample inputs for which the ion suppression effects were most severe (Fig. 4T and Supplementary Fig. 4).Common metabolites identified in both plasma and urine revealed that ion suppression correction yielded comparable results in the two matrices (Fig. 4U).Overall, the ability of the IROA Workflow to produce a linear increase in suppression-corrected signals consistently with increasing sample input demonstrates the broad applicability and versatility of the Workflow for ion suppression correction across diverse biofluids.Moreover, our results demonstrate that the Workflow successfully corrects ubiquitous ion suppression across diverse metabolic pathways, analytical techniques, and experimental conditions (Fig. 5A, B, & C and Supplementary Fig. 5).

Development of Dual MSTUS for robust normalization of metabolomic data.
The Workflow presented in the previous section promises to improve quantitative accuracy and correct for a subset of factors that introduce variation in metabolomic assay sensitivity, but ion suppression correction alone cannot enable accurate comparison of metabolite levels across variable sample input, analytical batches, biological matrices, instruments, and many other factors.Inconveniently, accurate crossmatrix comparison requires absolute quantitation using matrix-matched external calibration curves.Dual MSTUS normalization provides a solution to those challenges.MSTUS normalization determines a total useful signal value for each sample by excluding all clearly non-relevant signals and then sums the remaining data 22 .That relatively simple strategy has been successful in metabolomics.Here we extend the MSTUS strategy into a Dual MSTUS algorithm.Since the IROA-IS 13 C values for each metabolite should be equal across all samples, adjusting each MSTUS-12C value to the corresponding MSTUS-13C value normalizes all 12 C signals to produce accurate comparisons across variable conditions (Fig. 6).An important distinction here is that, unlike suppression correction, wherein each compound is corrected individually, MSTUS normalization calculates a Normalization Factor, NF, that is applied to all peaks in each sample (see Methods for detail).
When applied to the data presented in the previous section, Dual MSTUS normalization produced significantly improved precision across different chromatographic methods and conditions (Fig. 6A-K).Notably, metabolites lacking a matched IROA-IS could not be suppression-corrected but still benefited from Dual MSTUS normalization because NF is a sample-based value and not a metabolitespecific value (Supplementary Fig. 6 and Supplementary Table S7).Since those measurements were not corrected for ion suppression, the normalized data still reflected suppressed signal, but Dual MSTUS normalization improved the biological and statistical significance of comparisons across samples, analytical conditions, and matrices with dramatic decrease in error from 16% to below 1% in plasma and from 50+% to below 1% in urine (Fig. 6L and Supplementary Table S8).Note that we designed the experiment to include extreme variance in sample input (50 -1,500 µL plasma extract), which is not likely to occur in ordinary experiments.Under typical experimental conditions with the same chromatographic system and equal input across all samples, the Workflow including Dual MSTUS normalization can be expected to improve insight generation with greater biological significance and statistical significance.

Discussion
We present an IROA-based Dual MSTUS Workflow that is applicable across chromatographic systems, ion sources, and biological matrices.The Workflow, including suppression correction and Dual MSTUS normalization algorithms, facilitates accurate comparison of metabolomic profiling data across: i) unclean and clean ionization sources; ii) IC-, HILIC-, and RPLC-MS systems; and iii) plasma and urine matrices.Overall, the Workflow provides an almost universal antidote to the negative effects of ion suppression across all analytes in a non-targeted metabolite profiling study.By doing so, it can increase the sensitivity, accuracy, precision, and biological and statistical significance of such measurements.
The IROA Workflow has significant implications for the metabolomics field.Matrices like fineneedle aspirates and stool, which produce high, variable suppression can benefit most profoundly.Fine-needle aspirates (FNAs) are central to the diagnosis and management of breast cancer, and stool profiling is fundamental to microbiome metabolite profiling.Notably, uncleaned MS system components variably affect intra-and inter-instrument performance within and across laboratories, but the IROA Workflow produces equivalent data for equivalent samples processed across variable conditions.Since the very high attrition rate for clinical biomarker development is thought to be associated with hidden artifacts in the source data, inappropriate statistical methods, and unmanageable variation 29,30 , the improved rigor and reproducibility provided by the IROA Workflow could decrease metabolomics-based biomarker failure.
The Workflow is highly effective but does have limitations.Chief among them is that the output includes only metabolites that are detected in both the 12 C and 13 C channels.One factor that can limit such detection is instrument sensitivity.Another factor is ion suppression.A third factor is low abundance in the IROA-IS.Ongoing advances in MS sensitivity and in the IROA-IS itself will translate into increased coverage of the metabolome and increased utility of the IROA TruQuant workflow.
By addressing multiple sources of error, the IROA Workflow effectively corrects for a majority of the intrinsic inaccuracies of MS analyses and could help to move the metabolomics field toward standardization of non-targeted analyses.

Sample preparation
For each chromatographic system tested, triplicate 225 µL aliquots of human plasma with K2-EDTA as the anticoagulant (BioIVT lot #BRH1225104) were added to separate 15 mL polypropylene centrifuge tubes and combined with 5.4 mL dry ice-cooled methanol, and centrifuged at 16,100 g at 4 °C for 10 minutes.The supernatant was transferred to a new tube from which aliquots of 1500, 1250, 1000, 750, 500, 250, 100, and 50 µL were prepared.The methanol extracts were then dried using a centrifugal vacuum concentrator and stored at -80°C.Immediately before analysis, samples were reconstituted in 40 µL of IROA-IS solution, prepared by dissolving the contents of an IROA-IS vial in 1.2 mL H2O, vortexing, and briefly centrifuging.Following the IROA TruQuant protocol (Fig. 2), the entire contents of the tubes were then transferred to polypropylene autosampler vials.IROA-LTRS, an isotopicallylabeled complex standard composed of extracts universally labeled at both 5% and 95% U-13C in a 1:1 ratio, was prepared by dissolving the contents of an IROA-LTRS vial in 40 µL H2O.In addition to blank and process blank samples (deionized H2O, before and after SOP), injections of the 350 µL aliquot plasma extract sample (without IROA-IS) and the IROA-IS solution alone served as controls.A 3 µL injection of IROA-LTRS and 5 µL injection of each sample were analyzed.

Ultra-high resolution mass spectrometry
For HILIC and RPLC chromatographic systems, a Thermo Scientific Orbitrap Fusion Lumos Tribrid mass spectrometer was operated in full scan mode using a scan range of 70-1,400 m/z and a resolution of 240,000.Heated electrospray positive ionization used a spray voltage of 3,500 (V) and vaporizer and capillary temperatures of 250 and 375°C, respectively.The sheath, auxiliary, and sweep gas pressures were 35, 10, and 0 (arbitrary units), respectively.Identical parameters were used for negative ionization with the exception of -2,800 (V) spray voltage.For IC-MS, data were acquired using a Thermo Orbitrap IQ-X Tribrid Mass Spectrometer under ESI negative ionization mode with settings otherwise as noted above.

Data processing and calculations
Raw data files acquired by XCalibur software (Thermo Fisher) were converted to mzXML format using ProteoWizard's msConvert (version 3.0.193336,64 bit) using only the "peak picking by vendor" mode (both MS 1 and MS 2 scans were converted).The mzXML files were analyzed using ClusterFinder (version 4.2.21,64-bit, IROA Technologies).IROA-LTRS mzXML files were analyzed in non-targeted mode to determine the identity, RT, and MS 1 /MS 2 characteristics of the IROA-LTRS peaks.Those data were then used to create runtime libraries (Dual MSTUS Metabolite Lists), which were used as the basis for annotation and quantitative analyses of experimental samples.The quantitative data included raw, suppression-corrected, and normalized data for each peak in the analytical samples.Annotation data included the name of the compound, the reporting standard achieved, identity algorithm used (all inferred by comparison to the IROA-LTRS), retention time (RT), and m/z for both 12C and 13C monoisotopic peaks.
Calculations were performed automatically within ClusterFinder, using the algorithms described in the following sections.

Quality control and quality assurance
The following additional sample types were included in the analysis: a) a distilled water blank, b) a sample with an aliquot level of 350 µL extract that did not contain the Internal Standard mixture, and c) a sample that contained the Internal Standard mixture but none of the experimental extract.The TruQuant Protocol also uses IROA-LTRS as a Quality Assurance that instrument performance is within acceptable levels and to aid in peak annotation under current runtime conditions.The use of the LTRS and these additional sample types are further described in the Results and Discussion section and in Supplemental Materials.

Suppression-correction calculation
As part of the experimental design ion suppression was intentionally induced by increasing sample concentration.The following algorithm was employed for the correction of ion suppression.
In Equation 1, the raw MS data provide values for [AUC-12Csuppressed] and [AUC-13Csuppressed]. [AUC-13Csuppression-corrected] can be approximated from the least suppressed value in the data set-the Internal Standard Only (IS-ONLY), internal standard sample.For optimal accuracy, it is critical that the AUCs include the sum of all of the isotopologs derived from the sample (12C end of the isotopolog ladder) and the internal standard (13C end of the isotopolog ladder).
It is important to note that other values can be used in place of the "least suppressed value" to estimate [AUC-13Csuppression-corrected].For instance, if the concentration of the internal standard is held constant across experiments performed over a period of time, then the leastsuppressed value could be a historically and more accurately determined value.The leastsuppressed value could be either a quantitative molar quantity or a relative intensity (e.g., an AUC).The values determined from Equation 1 can be used to calculate % suppression as (1-[AUC-12Csuppressed]/[AUC-12Csuppression-corrected]).

Dual MSTUS algorithm
We developed a modified version of the mass spectrometry total useful signal (MSTUS) algorithm as follows 28 .Because it uses two MSTUS values, i.e. that of the IROA-IS sample ([MSTUS-13C]) and the original natural abundance sample ([MSTUS-12C]), we refer to this as a "Dual MSTUS algorithm" (see Results and Discussion).The Dual MSTUS Metabolite List for each sample includes only those peaks that are found in both the sample ( 12 C) and IROA-IS ( 13 C).
[MSTUS-12C] = the sum of the [AUC-12C suppression-corrected] values for all 12C peaks that are paired with a 13C peak.
[MSTUS-13C] = the sum of the [AUC-13C suppression-corrected] values for all IS peaks that are paired with a 12C peak.

Equation 2: NF = [MSTUS-12C] / [MSTUS-13C]
NF considers the total peak intensity in the sample relative to the total peak intensity in the IROA-IS to adjust the total content of the sample to the total content of the internal standard.Leveraging the statistical principal that variance decreases with increasing sample size, greater accuracy is achieved when a larger number of metabolites is used to establish the total content of each sample.
Step 3: Calculate normalized values for each metabolite in each sample.
Once NF has been determined for each sample, each [AUC-12C suppression-corrected] value (Equation 1) is normalized by dividing by NF as indicated in Equation 3. The calculated NF values were extremely close to the values expected from the known aliquot sizes (Fig. 6).

Equation 3: [AUC-12C normalized] = [AUC-12C suppression-corrected] / NF.
The process for Dual MSTUS normalization is straightforward.It involves two main steps.First, NF is computed for each sample (Equation 2) based on the Dual MSTUS Metabolite List, which includes peaks found in both the 12 C and 13 C channels of each sample.Second, the suppression-corrected 12 C values (Equation 1) of each metabolite in the Dual MSTUS Metabolite List are adjusted so that their sum is equal to the total [MSTUS-13C] value.That adjustment aligns with the original MSTUS rationale that normalizing over many metabolites will average out experimental variance, improving our ability to analyze and interpret the underlying biology accurately.By first suppression-correcting then applying the dual MSTUS NF, multiple sources of variability can be mitigated to provide greater rigor and reproducibility in non-targeted metabolomics.The total number of metabolites detected in each experimental condition determine the normalized linear dynamic range.

Fig. 1 |
Fig. 1 | Pre-analytical and analytical variables that affect quantitative rigor and reproducibility in MS-based metabolomics.The red line represents observed, suppressed signal, which can be caused by the listed factors.The green line represents suppression-corrected signal, which can be realized by the techniques described in this paper.Representative examples of factors that affect the accuracy of MS-based measurements are listed and have been reviewed elsewhere 12,13 .

Fig. 2 |
Fig. 2 | IROA TruQuant Workflow.In this protocol the experimental samples (A) are prepared and dried.They are then reconstituted with a solvent containing the IROA-IS (B) to yield the analytical samples (C).The analytical samples are randomized and injected within a sequence that starts and ends with injections of the IROA-LTRS (D), which is also injected approximately every 10 injections.Based on the presence of the IROA-IS, each sample can be suppression-corrected and normalized despite significant differences in sample input (original sample aliquot volume prior to dry down) (E).

Fig. 4 |
Fig. 4 | Ion suppression correction by IROA-IS across chromatographic systems and conditions.(A) The IROA ion suppression correction workflow.(B) Number of MSTUS peaks detected across analytical conditions.(C-K) Raw MSTUS-12C (blue lines) and suppression-corrected MSTUS-12C (red lines) values are shown for: (C) HILIC positive mode, uncleaned source; (D) HILIC positive mode, clean source; (E) HILIC negative mode, uncleaned source; (F) HILIC negative mode, clean source; (G) RPLC positive mode, uncleaned source; (H) RPLC positive mode, clean source; (I) RPLC negative mode, uncleaned source; (J) RPLC negative mode, clean source; and (K) IC negative mode, cleaned source.(L) Ratio of raw MSTUS-12C to suppression-corrected MSTUS-12C peak intensity across chromatographic methods and experimental conditions.(M) Raw and suppression-corrected tryptophan values in RPLC positive ionization mode with cleaned source.(N) Raw and suppressioncorrected pyroglutamic acid values in IC negative ionization mode.(O) Identified chemical composition in entire RPLC clean dataset, as an example.(P) Kohonen Self Organizing Maps (SOM) show suppression patterns in the RPLC Clean raw data set for all 539 compounds.(Q) Density map show compounds associated with each of the patterns discovered in Fig. O. (R) Raw MSTUS-12C (blue lines) and suppression-corrected MSTUS-12C (red lines) values are shown for RPLC positive mode and RPLC negative mode (S) for urine.(T) Ratio of raw MSTUS-12C to suppression-corrected 12C peak intensity for urine matrix in positive and negative ion modes.(U) Plasma and urine MSTUS-12C signals for 4 common metabolites before and after suppression correction.12C SC =12C suppression corrected MSTUS; 13C raw = 13C raw MSTUS.

Fig. 5 |
Fig. 5 | Global metabolic pathway analysis illustrating the effects of IROA ion suppression correction.(A) Raw and suppression-corrected (SC) peak intensities across chromatographic systems and uncleaned and cleaned ionization source conditions.(B) Metabolic pathways determined by IC-MS before and after ion suppression correction.(C) Metabolic pathways determined by RPLC-MS before and after ion suppression correction.Data were drawn in SBGN (system biology graphical notation), Process Description (PD) and Activity Flow (AF) languages or Simple Interaction Format (SIF).Metabolites are color-coded based on the percent peak intensity.Raw = 12C raw; SC= 12C suppression corrected; Uncleaned+ = Uncleaned positive ion mode; clean+ = clean positive ion mode; uncleaned-= uncleaned negative ion mode; cleaned-= cleaned negative ion mode.

Fig. 6 |
Fig. 6 | Dual MSTUS normalization standardizes metabolomic data across a broad range of analytical conditions.Raw MSTUS-12C values (blue lines), normalized MSTUS-13C values (green lines), and suppression-corrected MSTUS-12C values (red lines) for indicated chromatographic systems and conditions in plasma.(A and B) HILIC positive mode, uncleaned and cleaned source; (C and D) HILIC negative mode, uncleaned and cleaned source; (E and F) RPLC (C18) positive mode, uncleaned and cleaned source; (G and H) RPLC (C18) negative mode, uncleaned and cleaned source; (I) IC-MS negative mode, clean source.Dual MSTUS normalization for urine by RPLC uncleaned positive mode (J) and negative mode (K).(L) Percent coefficient of variation (%CV) for raw, suppression-corrected, and normalized data from uncleaned and cleaned source conditions across different sample matrices and chromatographic systems including IC, HILIC, and RPLC.