Comparative evaluation of in-depth mass spectrometry and Olink Explore 3072 for plasma proteome profiling

Comparative evaluation of in-depth mass spectrometry and Olink Explore 3072 for plasma proteome profiling | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Comparative evaluation of in-depth mass spectrometry and Olink Explore 3072 for plasma proteome profiling Noora Sissala, Haris Babačić, Isabelle R. Leo, Xiaofang Cao, Jenny Forshed, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6501601/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Nov, 2025 Read the published version in Communications Chemistry → Version 1 posted You are reading this latest preprint version Abstract Recent advances in proteomics technologies have expanded the depth and scale of plasma proteome analyses, offering new opportunities for biomarker discovery and precision medicine. However, understanding the strengths and limitations of these technologies is crucial for study design and platform selection. We evaluated the performance and quantitative agreement of an in-depth mass spectrometry method (HiRIEF LC-MS/MS) and the Olink Explore 3072 proximity extension assay on 88 plasma samples, with 1,129 proteins analyzed with both methods. The technologies demonstrated complementary proteome coverage, high precision, and concordance in differential abundance analysis. Quantitative agreement in protein levels was moderate (median correlation 0.59, inter-quartile range 0.33–0.75), influenced by various technical factors. In addition, we introduce PeptOlink, a public resource for analyzing peptide-level quantitative agreement. Our findings highlight the complementary strengths of mass spectrometry proteomics and Olink Explore 3072, underscoring the value of combining both for comprehensive and reliable profiling of the plasma proteome. Mass Spectrometry Analytical Biochemistry Applied Biochemistry Biochemical Research Methods Translational Medicine Personalized Medicine Plasma proteomics Mass spectrometry Affinity proteomics Proximity extension assay Olink Explore 3072 Biomarker discovery Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Precision medicine and precision health depend on biomarkers to guide and facilitate disease prevention, diagnosis, prognosis, and treatment in an increasingly personalized manner. Blood plasma is a promising source of potential biomarkers, as it can be sampled in a minimally invasive manner and contains valuable biological and physiological information from all the organs in the body, harbored in a complex mixture of molecules such as nucleic acids, lipids, metabolites, and proteins ( 1 ). Proteins can be especially useful as biomarkers, since their levels are closely associated with the phenotype of the individual. Proteins are the effector molecules of cells and tissues, and their levels vary in different physiological and pathological states in a way that cannot fully be predicted by the genome or transcriptome ( 2 ). The significance of proteins for clinical practice is demonstrated by the fact that around 40% of laboratory blood tests in the clinic measure proteins ( 3 ). While the plasma proteome holds great potential for biomarker discovery, analyzing it has proven challenging due to its high dynamic range of concentrations. Protein concentrations in plasma span at least 12 orders of magnitude. Among the thousands of proteins in plasma, the 22 most abundant constitute 99% of the plasma proteome by mass ( 4 ). Disease-related proteins are often present at low levels in plasma and their detection has historically required either extensive sample processing for untargeted analysis or targeted analysis of individual or small sets of proteins. However, recent advancements in both global mass spectrometry (MS) and highly multiplexed affinity-based proteomics techniques have alleviated this problem by simultaneously increasing proteome coverage and sample throughput ( 5 , 6 ). Consequently, larger cohorts can be profiled comprehensively, increasing the potential for insights into human health and disease and facilitating the discovery of new biomarkers. In global MS-based approaches, proteins are measured in an untargeted manner by digesting proteins into peptides, separating and ionizing the peptides, measuring their mass-to-charge ratios with MS, and identifying and quantifying the peptides by matching their mass spectra to theoretical mass spectra from sequence databases (peptide-spectrum matching). Generally, MS-based approaches offer highly specific identification and quantification of detected proteins (peptides) but require time-consuming sample preparation to profile the plasma proteome in depth. Thus, while MS-based approaches can reach an analytical depth of thousands of proteins, studies employing MS proteomics have been limited in their sample size compared to those employing affinity-based methods ( 7 – 10 ). In contrast, affinity-based approaches, where affinity molecules such as antibodies or aptamers are used to bind and quantify predefined target proteins, enable high-throughput profiling of the plasma proteome ( 11 ). These methods, including the antibody-based proximity extension assays (PEAs) from Olink and the aptamer-based SomaScan assays from SomaLogic, have facilitated large-scale studies involving thousands of individuals ( 12 , 13 ). However, unlike MS, these methods do not provide direct detection of proteins (peptides), and ensuring the specificity and accuracy of the affinity binders is a challenge. To help mitigate this issue, PEAs rely on two antibodies to detect each target protein ( 14 ). Understanding the relative strengths and limitations of the various technologies available for plasma proteome profiling is essential to allow informed decisions regarding study design, platform selection, and data quality control (QC). Numerous studies have compared the performance of Olink’s PEAs and SomaLogic’s SomaScan assays ( 15 – 23 ), but few have compared MS and Olink’s PEAs, and these comparisons have been limited in their profiling depth ( 22 , 24 , 25 ). Here we present a comprehensive comparative evaluation of the Olink Explore 3072 PEA-based platform and our previously published method for in-depth MS-based plasma proteomics, high-resolution isoelectric focusing coupled with liquid chromatography and tandem mass spectrometry (HiRIEF LC-MS/MS) ( 7 ). This workflow involves depletion of high-abundance proteins, tandem mass tag (TMT) labeling, extensive pre-fractionation of peptides using HiRIEF, and data-dependent acquisition (DDA) to achieve high analytical depth and relative quantification. We evaluate the two methods in terms of proteome coverage, precision, statistical power, and quantitative agreement at both the protein and peptide level. Finally, we present PeptOlink, a public resource for analyzing the agreement between peptide and protein abundances, as quantified by MS and Olink, respectively. Results Overview of the proteomics datasets We detected 2,578 unique proteins across 120 samples using HiRIEF LC-MS/MS (114 distinct samples with six samples run in duplicate) and measured 2,923 proteins in a subset of 88 samples using Olink Explore 3072 (Fig. 1). In the Olink data, Normalized Protein Expression (NPX) values below the limit of detection (LOD) were retained but considered missing values. Ten proteins with NPX values below the LOD in all samples were deemed not detected and were excluded from further analysis. In total, 4,362 proteins were detected and quantified in at least one sample across both technologies, 2,578 with MS and 2,913 with Olink, with 1,129 overlapping between methods (Fig. 2A, Table S1). The number of overlapping proteins varied by Olink Explore panel, with the greatest overlap observed for the Cardiometabolic panel (Fig. 2B). While both technologies detected a large number of proteins, there was a clear difference in the frequency of missing values. In the MS data, 55% of all quantified proteins had at least one missing value, compared to 20% of proteins in the Olink data (Fig. 2C). A total of 1,822 proteins were detected in at least 50% of the samples with MS, and 2,460 with Olink, while 1,212 and 1,910 were detected in all samples with MS and Olink, respectively. In the MS data, missing values were TMT set-specific (Figure S1). Plasma proteome coverage To compare the plasma proteome coverage of each method, we curated a reference set of 4,889 plasma proteins by compiling proteins from the PeptideAtlas (build 2023-04, www.peptideatlas.org) (26) and the Human Protein Atlas (HPA, www.proteinatlas.org) (27, 28) (see Methods). HiRIEF LC-MS/MS showed a greater overlap with this reference plasma proteome, while Olink Explore 3072 targeted more than a thousand proteins not reported in the MS-based studies found in the PeptideAtlas (Fig. 2D, Table S1). Combined, the platforms covered 63% of proteins included in the reference plasma proteome. Next, we evaluated plasma proteome coverage across concentration ranges, based on the estimated concentration in blood of each protein from the HPA (28). Both technologies detected proteins with estimated plasma concentrations spanning 10 orders of magnitude, down to the pg/mL concentration range (Fig. 2E). However, low-abundance proteins frequently had a large proportion of missing values, especially in the MS data (Fig. 2F). Olink demonstrated a higher coverage of low-abundance proteins, while MS demonstrated a higher coverage of mid to high-abundance proteins (Fig. 2G). Therefore, proteins detected exclusively by Olink were mainly low abundance, whereas those detected exclusively by MS tended to have higher concentrations (Fig. 2E). These observations remained consistent when considering proteins detected in at least 50% of samples, although the coverage of low-abundance proteins decreased for both methods (Figure S2). Characterization of detected proteins To further characterize the types of proteins that could be measured with each technology, we compared the frequency of various annotations from the HPA, performed overrepresentation analyses (ORA) of gene ontology (GO) terms, and investigated the coverage of plasma protein biomarkers approved by the United States Food and Drug Administration (FDA). In this cohort, predicted secreted proteins, enzymes, metabolic proteins, immunoglobulins, proteins enriched in liver tissue, and potential drug targets were more frequent in the MS data, while predicted membrane proteins, CD markers, proteins secreted in the male reproductive system, and proteins enriched in the brain and testis were more frequent in the Olink data (Fig. 3A, Table S2). It should be noted that 95 proteins (3.7%) detected by MS were not found in the HPA, compared to only 22 proteins (0.76%) detected by Olink, which could slightly bias the comparison of annotation frequencies between platforms in favor of Olink. Consistent with the observation that MS detected proportionally more mid to high-abundance plasma proteins, MS was enriched for GO biological processes inherent to the plasma proteome—hemostasis, blood coagulation, complement activation, and metabolism (Fig. 3B, Table S3). In contrast, Olink was enriched for GO biological processes related to low-abundance signaling proteins, particularly cytokines. The two methods detected comparable numbers of FDA-approved plasma protein biomarkers (29)—74 (MS) and 72 (Olink) out of 99, with 55 biomarkers detected by both (Table S4). Biomarkers exclusively detected by MS included various transport and metabolic proteins, whereas Olink exclusively covered several hormones. Precision To evaluate the precision of repeated measurements with HiRIEF LC-MS/MS and Olink Explore 3072, we calculated technical coefficients of variation (CV) for each protein across duplicate samples (Table S5). For Olink, intra-assay CVs were computed for 1,797 proteins (62%) after excluding values < LOD, based on a control sample of pooled donor plasma run in duplicate on the same plate. For MS, inter-assay CVs were calculated using duplicates of patient samples, with each replicate run in a different TMT set. Due to variations in protein identifications between TMT sets ( i.e. , missing values), CVs could be calculated for 1,952 proteins (76%). Both MS and Olink demonstrated high precision, as indicated by low technical CVs (Fig. 4A). The median CV was slightly lower for Olink than for MS (5.7% vs. 6.8%, p = 1.7 × 10⁻¹⁶). Most proteins had a CV below 15% in both datasets—85% in the MS data and 80% in the Olink data—although Olink had a higher proportion of proteins with very low CVs, below 5% (40% vs. 35%). However, it is important to note that the Olink CVs were intra-assay CVs, while the MS CVs were inter-assay CVs. Therefore, the Olink CVs may have been underestimated. Technical CVs were higher for proteins with more missing values, as well as for proteins with lower estimated concentrations in the blood (Figure S3). Statistical power Next, we investigated how the precision of HiRIEF LC-MS/MS and Olink Explore 3072 affected the statistical power of the differential abundance analysis (DAA) between males and females in both datasets. To ensure an equal number of tests for both technologies and equal sample sizes in each test, we focused the analysis on proteins quantified with both methods with no missing values (N = 569). After correcting for multiple hypothesis testing, there were 82 (14%) and 118 (21%) differentially abundant proteins (DAPs) between the sexes in the MS and Olink data, respectively (Table S6). Of these, 53 (9%) were statistically significant in both datasets (Fig. 4B). Therefore, 65% of DAPs found with MS were also found with Olink, while only 45% of DAPs found with Olink were also found with MS. These observations suggest that Olink may have had higher statistical power in detecting DAPs, whereas DAPs identified with MS were more likely to be reproduced with Olink. The platforms agreed on the direction of the difference for most proteins (80%), regardless of statistical significance. When considering proteins significant in at least one platform, the agreement was close to perfect (95%) (Fig. 4C). Thus, although the two methods had strong concordance in estimating differences in protein levels between groups, only a portion of the differences were consistently statistically significant. Finally, we investigated the effect of technical and biological variance on the statistical significance of sex-related differences in protein levels. The technical CVs in MS were slightly higher for DAPs identified only with Olink (Figure S4A), suggesting that technical noise in MS may have masked some protein level differences. Overall, the log2-fold change estimates of DAPs had higher absolute values in the Olink data (Figure S4B). DAPs identified exclusively by Olink had lower estimated concentrations in the blood than the rest of the DAPs (Fig. 4D). These findings suggest that Olink may have quantified some low-abundance proteins with higher precision, contributing to a larger number of DAPs. On the other hand, there were still many proteins, generally higher in abundance, that were detected as DAPs only with MS. Cross-platform correlation of protein levels To investigate the accuracy of protein identifications and quantifications, we estimated the quantitative agreement between HiRIEF LC-MS/MS and Olink Explore 3072 for each protein using Spearman’s rank correlation coefficient. On the complete dataset of overlapping proteins (N = 1129), the median correlation between paired MS and Olink protein measurements was ρ = 0.59, with nearly two thirds exhibiting a moderate to strong correlation of ρ ∈ [0.5,1.0] (Fig. 5A, Table S7). Several proteins had near perfect agreement, with the highest correlations observed for MBL2, PZP, ANGPT1, MYL3, DPT, and SHMT1 (ρ ∈ [0.95,0.97]). In contrast, some proteins had strong disagreement, with negative correlation, for example PAXX, SRPK2, GLIPR1, IL10RB, ISM2, and LSM1 (ρ ∈ [-0.71, -0.40]). Proteins with very low cross-platform correlations generally had many missing values (Fig. 5B). The correlations improved slightly after removing values < LOD and QC warnings in the Olink data (N = 1064), and further after additional filtering for proteins with a maximum of 50% missing values or QC warnings (N = 791) (Figure S5, Table S7). On the subset of overlapping proteins with no missing values or QC warnings (N = 463), the median correlation reached ρ = 0.68, with 81% of proteins having a moderate to high correlation between platforms (Fig. 5C, Table S7). Among proteins with no missing values, the lowest correlations were found for PPP1R9B, GDF2, SEMA3G, CTSL, ADGRF5, DAG1, and AGT (ρ ∈ [-0.23, -0.05]). Comparison with previous studies Next, we compared the correlations between HiRIEF LC-MS/MS and Olink Explore 3072 measurements with those from previous studies that assessed the agreement between MS, Olink, and/or SomaScan platforms (see Methods). Most previous studies have focused on the agreement between Olink and SomaScan (15–22). On average, our estimates of MS-Olink correlations were higher than the Olink-SomaScan correlations reported in most of these studies (Figure S6, Table S8). In a recent large-scale comparison of Olink and SomaScan platforms, Eldjarn et al . (15) categorized proteins into three confidence tiers according to the reliability of their measurements, with tier 1 representing highest confidence, and tier 3 lowest confidence. The classification was based on cross-platform correlations and the presence of protein quantitative trait loci (pQTLs) (see Methods). To provide additional orthogonal validation with MS, we examined our estimates of MS-Olink correlations in the context of these confidence tiers. Tier 1 proteins had a clearly higher median correlation between HiRIEF LC-MS/MS and Olink Explore 3072 (ρ = 0.72), compared to tier 2 (ρ = 0.53) and tier 3 (ρ = 0.47) proteins (Fig. 5D, Table S9). This supports the idea that the presence of pQTLs on both platforms, along with a high cross-platform correlation, is indicative of more accurate protein quantification. The median correlation between HiRIEF LC-MS/MS and Olink Explore 3072 for all confidence tiers were higher than the corresponding Olink-SomaScan correlations (Figure S7). However, the difference was most pronounced for tier 3 proteins, where the median HiRIEF-Olink correlation was ρ = 0.47, compared to ρ = 0.05 for Olink-SomaScan. These results suggest that the Olink assays for tier 3 proteins may be more accurate than previously indicated. Overall, our data provide orthogonal validation for the quantification accuracy of many Olink, and by extension SomaScan assays, in tier 1, and for a few assays in tier 3, based on strong cross-platform correlations in both studies (ρ > 0.7, Table S9). To date, only a few studies have reported correlations between MS and Olink measurements, with a limited number of overlapping proteins (22, 24, 25). These studies have shown varying levels of agreement between MS and Olink, with median correlations ranging from ρ = 0.27 to ρ = 0.56 for proteins overlapping with the present study (Figure S8, Table S10). On average, our MS-Olink correlations were higher than those of previous studies for the corresponding proteins. One of these studies, by Dammer et al. (22), also assessed the agreement between MS and SomaScan measurements. Our MS-Olink correlations were comparable to the MS-SomaScan correlations reported in their analysis (Figure S9, Table S11). Technical factors affecting cross-platform correlations To understand the poor quantitative agreement observed for some proteins, we investigated the associations between the MS-Olink correlations and various data quality indicators (Table S6). First, we found that lower correlations were linked to more missing values in both platforms (Fig. 6A-B, Figure S10). The proportion of missing values had a very minor and inconsistent effect on the spread of the data, making it unlikely that the association between the missing values and the cross-platform correlations were driven by the spread of the data itself (Figure S11). The higher percentage of missing values likely indicated noisier quantification, as proteins with more missing values had lower estimated concentrations, higher technical CVs, and median values closer to the LOD in the Olink data (Figure S12). Consequently, all these factors were associated with weaker cross-platform correlations (Figure S13). Second, cross-platform correlations were positively associated with the number of peptide spectrum matches (PSMs) and peptides used for quantification by MS, as well as with sequence coverage (Fig. 6C-D, FigureS13). Proteins with fewer PSMs, peptides, and lower sequence coverage also tended to have more missing values, further explaining the observed relationship between missing values and weaker correlations (Figure S14). The correlations were lower for proteins with only one median PSM or peptide, but still moderate on average (median ρ = 0.42, Figure S15), suggesting that a low number of PSMs or peptides alone is not sufficient to deem a protein quantification unreliable. Additionally, proteins with high precursor mass errors (> 5) in the MS data had weak correlations with Olink (Figure S16A). Most, but not all, of these proteins also had a large proportion of missing values in MS (Figure S16B). In summary, while missing values seem to be the strongest data quality indicator for MS, factors such as PSMs, peptides, and precursor mass errors can provide additional insights and further explain variations in MS-Olink correlations. Finally, we investigated how correlations were affected by QC warnings and differed between the Explore panels in the Olink data. Although MS-Olink correlations were lower among values with a sample QC warning, they only affected a small proportion of the proteins quantified in each sample (Figs. 6E and S17). Only 30 proteins that overlapped with MS had an assay QC warning and showed no difference in cross-platform correlation compared to the rest (Fig. 6E). The cross-platform correlation varied by Olink panel, likely driven by missing values, since panels with lower median correlation had a higher average proportion of missing values per protein (Figure S18A-B). Similarly, compared to version I panel proteins, correlations were somewhat lower for the version II panel proteins, which also had more missing values on average (Figure S18C-D). Thus, the impact of Olink data quality on the cross-platform correlations was likely largely driven by protein detectability, with a smaller effect from sample QC warnings. Protein properties affecting cross-platform correlations Next, we explored potential similarities in protein properties among proteins with poor correlations between MS and Olink. We found no difference in cross-platform correlations based on protein mass, length, or the number of isoforms reported in the UniProt database (Figure S19), and no enriched GO, MSigDB, KEGG, or Reactome gene sets by Gene Set Enrichment Analysis (GSEA) or ORA. However, enzymes, enzyme inhibitors, predicted secreted proteins, proteins secreted to the digestive system and the blood, as well as candidate cardiovascular disease genes from the HPA were enriched among proteins with high cross-platform correlations (Fig. 6F, Table S12). Most of these annotations were also overrepresented in the strong correlation group (ρ ∈ [0.7,1.0]) by ORA, while the no correlation group (ρ ∈ [-1,0.3)) had an overrepresentation of proteins related to intermediate filaments, mainly keratins (Table S13). These findings could be explained by the higher abundance of the proteins with strong MS-Olink correlations, while keratins could reflect sample contamination. Cross-platform correlations at the peptide level Lastly, we examined cross-platform correlations between HiRIEF LC-MS/MS and Olink Explore 3072 at the peptide level, to identify protein sequences or regions with varying correlations between platforms. We calculated the correlation between protein measurements of Olink assays and corresponding peptide measurements in the MS data, matched by gene name. To obtain more robust correlation estimates, we excluded peptides quantified in less than 15 samples with MS, resulting in a dataset of 13,856 peptides mapping to 822 genes and 847 unique UniProt IDs (Table S14). To enhance the accessibility of the results, we developed PeptOlink, a publicly available interactive R Shiny app (https://peptolink.serve.scilifelab.se/app/peptolink). PeptOlink allows users to visualize peptides quantified by MS on the protein sequence, along with their correlation with the corresponding Olink assay. It also provides filtering options for various peptide- and protein-level data quality factors and cross-platform correlation metrics, to aid in identifying proteins of interest. Below, we provide a few representative examples to illustrate the utility of PeptOlink in exploring cross-platform correlations: AMBP (α-1-Microglobulin/Bikunin Precursor), HYOU1 (Hypoxia Upregulated Protein 1), and MASP1 (Mannan-binding lectin Serine Protease 1). These proteins exhibited substantial variation in peptide-Olink correlations across different regions of their respective protein sequences. The AMBP gene encodes a precursor protein that is cleaved into α1-Microglobulin (residues 20–203) and Inter-α-Trypsin (IαI) Inhibitor Light Chain (residues 206–352). Notably, peptides mapping to the α1-Microglobulin region showed stronger correlations with the AMBP Olink assay (median ρ = 0.53) compared to peptides mapping to the IαI Light Chain region (median ρ = 0.07) (Fig. 7A, Figure S20), suggesting that the Olink assay primarily measures α1-Microglobulin. For HYOU1, we observed two regions with differing cross-platform correlations. One region, shared between isoforms 1 and 2 (residues 88–602), demonstrated poor agreement between MS and Olink measurements (median ρ = 0.21), whereas the other, unique to isoform 1 (residues 647–999), exhibited moderate agreement (median ρ = 0.64) between MS and Olink measurements (Fig. 7B, Figure S20). These findings suggest that MS and Olink may have detected different isoforms of HYOU1. Finally, MASP1 had multiple isoforms in the MS data (UniProt IDs: P48740, P48740-2, P48740-3) and evidence of differing MS-Olink correlation between these isoforms. At the protein level, isoforms 1 and 3 had poor correlation with the corresponding Olink assay (ρ = 0.04 and ρ = 0.23), while isoform 2, also known as MASP3, had a moderate correlation with Olink (ρ = 0.57). At the peptide level, regions mapping uniquely to isoform 2 exhibited higher cross-platform correlations than peptides mapping to other regions (Fig. 7C, Figure S20). These findings suggest that Olink’s antibodies may be binding sequences specific to isoform 2 of MASP1. In summary, these examples illustrate how peptide-level analysis by MS can reveal potential differences in proteoform measurements across proteomics platforms. Discussion In this study, we present a thorough investigation into the relative strengths and weaknesses of HiRIEF LC-MS/MS and the antibody-based Olink Explore 3072 platform, based on an in-depth analysis of the plasma proteome involving 4,362 proteins in total and 1,129 proteins overlapping between the two methods—significantly more than in previous studies comparing MS and Olink’s PEAs ( 22 , 24 , 25 ). We show that these platforms exhibit highly complementary proteome coverage, high precision, concordance and complementarity in DAA, and largely good, albeit variable, quantitative agreement in protein levels. Consequently, we provide orthogonal validation for a large proportion of the targets of the Olink Explore 3072 platform. Furthermore, we identify technical factors and protein properties influencing the quantitative agreement and compare estimates of cross-platform agreement across previous studies. Finally, we demonstrate how a peptide-level analysis of cross-platform correlations can reveal insights into differences in proteoform measurement. Our findings demonstrate the complementarity of in-depth MS and Olink Explore 3072 in providing a more comprehensive analysis of the plasma proteome than either approach on their own. This is in line with previous reports from studies comparing other MS workflows and the smaller Olink Target panels ( 24 , 25 ). Together, these technologies offer broad proteome coverage across a wide concentration range and distinct biological processes, increasing the potential for biologically and clinically relevant discoveries. The differences in proteome coverage suggest that each platform may offer distinct advantages in specific study contexts. For example, the relatively higher coverage of dedicated plasma proteins and metabolic proteins observed for MS may be advantageous for investigating cardiovascular disease, metabolic disorders, or any condition with large effects on the plasma proteome, such as systemic infectious or inflammatory diseases. Olink’s various Explore panels, which target many low-abundance tissue-leakage and signaling proteins, may offer an advantage in the study of conditions affecting specific tissues or the immune system, for example cancer, neurological disorders, or chronic inflammatory diseases. Beyond proteome coverage, these technologies have additional complementary strengths and limitations. The Olink Explore platform currently offers higher sensitivity and throughput than MS, enabling large-scale studies with high proteome coverage, though limited to pre-defined proteins. In contrast, the untargeted nature of MS provides a significant advantage for discovery-focused studies, allowing for a general characterization of the plasma proteome across different conditions and for the detection of novel proteins and PTMs. This advantage will likely become even more pronounced as sensitivity and proteome coverage continue to improve. In particular, the Orbitrap Astral has shown promising results in plasma proteome profiling, providing higher sample throughput while maintaining or even increasing profiling depth ( 30 , 31 ). Ultimately, the choice of profiling platform should be guided by the specific goals and disease context of the study. Since both methods provide valuable insights into different, although somewhat overlapping, aspects of the plasma proteome, combining in-depth MS with Olink panels targeting relevant low-abundance proteins offers an attractive approach to plasma proteome profiling that leverages the strengths of both technologies and can be tailored to the specific application. The complementarity of in-depth MS and Olink Explore 3072 was further reflected in the analysis of protein level differences between males and females. A greater number of DAPs were detected when combining both platforms, with DAPs unique to each platform stemming from different concentration ranges of the plasma proteome. The larger log2-fold change values and the greater number of DAPs identified by Olink, particularly in the low-abundance concentration range, likely reflect the sensitivity of the PEA technology. At these low concentrations, MS might be nearing its LOD, where the biological signal could be masked by technical noise. In addition, TMT ratio compression, an artifact caused by peptide co-fragmentation in MS experiments using TMT labeling ( 32 ), may have slightly reduced the observed differences in protein abundance between samples. Regardless of the underlying cause, larger sample sizes may be needed to reach statistical significance with MS, at least for some low-abundance proteins. Despite these limitations, MS identified numerous DAPs that were not detected by Olink, particularly in the high-abundance concentration range. In addition, the DAA exemplified how the agreement between platforms can be used to assess the validity of findings when using different profiling technologies. While only a subset of DAPs reached statistical significance with both methods, there was strong concordance in the estimated differences. Thus, because discrepancies in statistical significance could be attributed to differences in power, such as higher noise or more missing values in one platform, examining the agreement in protein level differences provides a more comprehensive evaluation of the results beyond just the significant proteins. Overall, this analysis illustrates the value of combining MS and Olink for enhancing the biological insights and increasing the number of potential biomarkers drawn from proteomic data. We identified several technical factors associated with weaker agreement between platforms, which can serve as indicators of lower measurement accuracy. These included a higher proportion of missing values, lower estimated protein concentrations, and higher technical CVs for both platforms, as well as sample QC warnings and median NPX values closer to the LOD for Olink data, and few PSMs or peptides and high precursor mass errors for MS data. In this study, we treated missing values in MS and values < LOD in Olink as equivalent, but it is important to recognize that these arise through different mechanisms. In MS workflows that employ TMT labeling, missing values are largely TMT set-specific, with peptide detection varying across runs. Detection in each run depends on a variety of biological and technical factors, not solely protein abundance or absence ( 33 ). In contrast, values < LOD in Olink assays more certainly indicate that the protein concentration was too low for reliable detection (or absent). Despite these differences, missing values had the strongest effect on lowering cross-platform correlations in both datasets, and proteins with a high proportion of missing values should therefore be handled cautiously in both types of proteomic data. Furthermore, we observed that some proteins with seemingly high-quality quantification on both platforms had poor agreement between platforms. This could be explained by a multitude of factors: the measurement of different isoforms, differences introduced by sample handling or preparation, antibody cross-reactivity, PTMs, and many more. Through an analysis of cross-platform correlations at the peptide level, we were able to identify potential differential detection of proteoforms and proteolytic cleavage products. At the protein level, the examples we presented, AMBP and HYOU1, had moderate correlation between platforms. However, our results from the peptide-level analysis demonstrate that the correlations could be stronger if both platforms were measuring the same protein regions or isoforms. These examples highlight an ongoing challenge in proteomics: shifting from a gene- or protein-centric to a proteoform-centric analysis of the proteome. While MS offers some ability to distinguish between proteoforms, there is still a need to better resolve the specific epitopes and isoforms targeted by affinity binders. Previous studies have made significant efforts to validate antibody specificity and selectivity using MS ( 34 – 36 ), and similar work will be needed to evaluate the affinity binders used in commercial plasma proteomics platforms such as Olink and SomaScan. Orthogonal validation with MS, as performed in the present study, represents one important pillar of affinity binder validation ( 37 ). The cross-platform correlation analysis revealed comparable but slightly higher correlations between MS and Olink than have been reported in previous studies ( 22 , 24 , 25 ), which could be explained by differences between MS workflows or differences between the Olink Target and Olink Explore panels. Our data provide evidence for the accuracy and specificity of at least one third of the Olink Explore 3072 assays with a strong correlation with HiRIEF LC-MS/MS. We also confirm with MS that the association with genetic variation—the presence of pQTLs—is useful to assess the accuracy of affinity-based assays. However, we also show that the pQTL approach is not without limitations, as proteins with pQTLs on both affinity-based platforms did not necessarily have a strong correlation between each other or with MS. The pQTL approach may penalize proteins that exhibit high cross-platform correlations but lack pQTLs due to an absence of genetic associations or small effect sizes. We identified a few such cases in the study by Eldjarn et al. ( 15 ), where the accuracy of the affinity-based assays was supported by our MS data despite a lack of pQTLs. Finally, through comparison with previous studies investigating cross-platform correlations between Olink and SomaScan platforms ( 15 – 22 ), our data may also indirectly provide support for the specificity of some SomaScan assays. However, due to the lack of SomaScan data in this study, we cannot draw definitive conclusions. Some limitations of this study should be considered. First, the CV calculations were based on a small number of duplicate samples, leading to uncertainty in the CV estimates. Additionally, we were unable to calculate intra- and inter-assay CVs for both technologies, and therefore, we compared the inter-assay CVs of HiRIEF LC-MS/MS to the intra-assay CVs of Olink Explore 3072. Second, some conclusions drawn from this study may not be fully generalizable to all other MS workflows and Olink platforms. Results could vary due to differences in sample preparation protocols, instruments, and data processing of MS workflows, or due to differences in affinity probes and performance characteristics of Olink platforms. These variations are particularly likely to affect the quantitative agreement between MS and Olink, as seen in the comparison of cross-platform correlations between studies. Finally, some results could differ between patient cohorts, due to the influence of factors such as age, sex, and disease state on the plasma proteome. Nevertheless, the main conclusions regarding the relative strengths and limitations of in-depth MS workflows and Olink platforms are likely to remain consistent. Future studies should include additional technical replicates and compare other in-depth MS workflows to affinity-based profiling platforms, both Olink and SomaScan, to further evaluate these technologies. In conclusion, this study provides insights into the complementary strengths and limitations of in-depth MS and Olink Explore 3072 for plasma proteome profiling. Our analysis, encompassing a large number of overlapping proteins, a thorough investigation into factors affecting platform performance, and an analysis of cross-platform correlations at both the protein and peptide level demonstrate that overall, both platforms perform well at profiling the proteome in depth, with each focusing on different portions of the plasma proteome. The complementary proteome coverage and the variable quantitative agreement highlight the added value of combining these platforms for a more comprehensive and reliable profiling of the plasma proteome, enabling a broader characterization of different diseases and more robust biomarker discovery. Materials and methods Ethical approval This study was approved by the regional ethical review board in Stockholm, Sweden (EPN: ref no 2014/1290–32) and conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent. Study design and sample collection The present study cohort consists of a retrospectively selected subset of the PEX-LC cohort, which has been described previously ( 38 ). Briefly, plasma samples were collected from patients referred to the Karolinska University Hospital (KUH) in Stockholm, Sweden, for investigation of suspected lung cancer between September 2014 and November 2015. The plasma samples were collected during the participants’ first visit to KUH, before diagnosis and treatment. Blood was drawn into EDTA tubes, centrifuged at 2500×g at RT for 10 min, and the resulting plasma samples were biobanked and stored at -80℃. The present study cohort includes 114 patients, with an equal number of patients diagnosed with either lung cancer or no cancer, i.e. other benign lung conditions. Samples from all 114 patients were analyzed using MS, and a subset of 88 plasma samples with Olink Explore 3072. For the MS analysis, six samples were run in duplicate (aliquoted at the start of sample preparation), resulting in a total of 120 samples. Plasma proteome profiling Mass spectrometry-based proteome profiling Plasma depletion and in-solution digestion To reduce sample complexity and increase the number of protein identifications, the 14 most abundant plasma proteins were depleted from the samples using High Select Top14 Abundant Protein Depletion Mini Spin Columns (Themo Scientific). 10 μL of plasma was applied to each column, the columns were incubated at room temperature for 20 min with gentle end-over-end mixing, and depleted flowthroughs were obtained through centrifugation. The sample buffer was exchanged to 50 mM HEPES (pH 7.6) using 5 kDa spin concentrators (5K MWCO, 4 mL, Agilent Technologies, 5185-5987) by centrifuging three times at 5000 rpm for 30 min. Protein concentration was measured using the Micro BCA Protein Assay Kit (Thermo Scientific, 23235) to estimate the total protein amount per sample. Next, proteins were digested into peptides using lysC and trypsin (sequencing grade modified, Pierce) following a previously described in-solution digestion protocol ( 39 ). In brief, 40 μg of protein from each sample was alkylated with 8 mM chloroacetamide. 100 μL of lysC buffer (0.5 M Urea, 50 mM HEPES, pH 7.6 and 1:50 enzyme-to-protein ratio) was added and the samples were incubated overnight. The same procedure was repeated for trypsin; 100 μL of trypsin buffer (50 mM HEPES, pH 7.6, 1:50 enzyme-to-protein ratio) was added and the mixtures were incubated overnight. Finally, the samples were dried in a SpeedVac and resuspended in 50 μL TEAB pH 8.5 to a final concentration of 100 mM. TMT labeling 40 μg of peptides from each sample was labelled with isobaric TMTs (TMTpro 16plex Label Reagent Set, Thermo Scientific) according to the manufacturer’s protocol. A total of 120 samples (114 distinct plasma samples, and six pairs of technical replicates) were labeled with eight sets of TMTpro 16plex, with one internal standard per set. The master pool of internal standards was made by pooling a small amount of protein from each sample, and the master pool was then split into eight internal standards of 40 μg of protein each. The TMT labeling scheme is shown in Table S15. The TMT labeling efficiency was determined by LC-MS/MS prior to pooling of the samples. For this, 1 μL of each sample of a TMT set was mixed, dried down, and resuspended in 10 μL of mobile phase A. Approximately 2 μg was injected into the LC-MS/MS system and analyzed with a 3h gradient. After confirming a labeling efficiency >95%, samples of the same TMT set were pooled. The eight resulting TMT pools were purified through solid phase extraction using SPE strata-X-C columns (Phenomenex), and purified samples were dried in a SpeedVac. High-resolution isoelectric focusing To further reduce sample complexity, pooled samples were pre-fractionated using HiRIEF, following a previously described protocol ( 40 ). Briefly, peptides were separated by their isoelectric point through immobilized pH gradient isoelectric focusing (IPG-IEF) on gel strips with a 3–10 pH gradient. After IEF, each gel strip was split into 72 fractions, and proteins from each fraction were eluted and transferred to a 96-well microtiter plate using a liquid-handling robot (GE Healthcare prototype). Finally, the fractionated samples were dried in a SpeedVac and stored at -20℃ until analysis with LC-MS/MS. LC-MS/MS Online LC-MS was performed as previously described ( 40 ) using a Dionex UltiMate™ 3000 RSLCnano System coupled to a Q-Exactive-HF mass spectrometer (Thermo Scientific). The contents of each plate well were dissolved in 20 uL of solvent A and 10 uL was injected. Samples were trapped on a C18 guard-desalting column (Acclaim PepMap 100, 75μm x 2 cm, nanoViper, C18, 5 µm, 100Å), and separated on a 50 cm long C18 column (Easy spray PepMap RSLC, C18, 2 μm, 100Å, 75 μm x 50 cm). The nano capillary solvent A consisted of 94.9 % water, 5 % DMSO, and 0.1 % formic acid, and solvent B consisted of 5 % water, 5 % DMSO, 89.9 % acetonitrile, and 0.1 % formic acid. At a constant flow of 0.25 μL/min, the curved gradient went from 6-10 % solvent B up to 40 % solvent B in each fraction in a dynamic range of gradient length (see Table S16), followed by a steep increase to 100% solvent B in 5 min. FTMS (Fourier transform mass spectrometry) master scans with 60 000 resolution and mass range 300-1500 m/z were followed by data-dependent MS/MS with a resolution of 30 000 on the top 5 ions using higher energy collision dissociation (HCD) at 30% normalized collision energy. Precursors were isolated with a 2 m/z window. Automatic gain control (AGC) targets were 1 6 for MS1 and 1 5 for MS2. Maximum injection times were 100 ms for MS1 and 400 ms for MS2. The entire duty cycle lasted ~2.5 s. Dynamic exclusion was used with 30 s duration. Precursors with unassigned charge state or charge state 1 were excluded. An underfill ratio of 1% was used. Protein identification and quantification Orbitrap raw MS/MS files were converted to mzML format using msConvert from the ProteoWizard tool suite ( 41 ). Spectra were searched using the ddamsproteomics Nextflow (v22.10.5) ( 42 ) pipeline (https://github.com/lehtiolab/ddamsproteomics, v2.11), which runs MSGF+ (v2020.03.14) ( 43 ) and Percolator (v3.04.0) ( 44 ) for peptide identification. All searches were performed against a database of all human proteins from the UniProtKB/Swiss-Prot release of May 2022. MSGF+ settings included precursor mass tolerance of 10 ppm, fully tryptic peptides, a maximum peptide length of 50 amino acids and a maximum charge of 6. Fixed modifications included carbamidomethylation on cysteine residues and TMTpro 16plex on lysine residues and peptide N-termini. A variable modification was used for oxidation on methionine residues. PSMs found at 1% false discovery rate (FDR) were used to infer protein identities. TMTpro 16plex reporter ions were quantified using OpenMS project's IsobaricAnalyzer (v2.5.0) ( 45 ). Relative quantification was calculated on peptide and protein level based on PSMs mapping to only one protein group (UniProt ID) and with 1% FDR. PSMs with missing values in any channel within a TMT set were excluded. Relative quantification values were calculated for each TMT channel as the median of PSM ratios (channel/internal standard). To obtain these PSM ratios, PSM intensities were log2-transformed, and the transformed PSM intensities of the internal standard were subtracted from the transformed PSM intensities of the channel. Peptide/protein quantification values were then normalized by subtracting the median of the channel from each value. Protein FDRs were calculated using the picked-FDR method using UniProt IDs as protein groups and limited to 1% FDR. Antibody-based proteome profiling The samples were analyzed using the Olink Proteomics PEA Explore 3072 at SciLifeLab Affinity Proteomics unit at Uppsala University and the National Genomics Infrastructure Uppsala. The detailed protocol for PEA Explore has been previously described by Wik and colleagues ( 14 ). In summary, Olink’s PEA Explore technology utilizes pairs of antibodies conjugated with single-stranded DNA oligonucleotide reporter molecules, known as probes, which bind to their respective targets if present in the sample. When both probes in a pair bind to their target in proximity, double-stranded DNA amplicons are generated. The Explore 3072 assay comprises eight distinct 384-plex panels targeting inflammation, oncology, cardiometabolic, and neurology proteins, covering a total of 2,923 human proteins. Four of these are version I panels ( e.g. Inflammation), which were part of the earlier Olink Explore 1536 platform, while the remaining four are version II panels ( e.g. Inflammation II), added to the Olink Explore 3072 platform. Following the initial probe-based immune reaction step in the PEA Explore workflow, the amplicons were extended and amplified in a two-step process, with individual sample index sequences added during the second step. After pooling the samples, the libraries were prepared and sequenced on a NovaSeq 6000 instrument (Illumina, San Diego, CA, USA). The raw BCL files were converted into count files, which were then translated into NPX values through a QC and normalization process incorporating internal and external controls, as specified by the manufacturer. In this process, QC is performed for each assay (protein) measured in each sample and for each assay overall. If the QC for an assay in a specific sample fails, the measurement receives a sample QC warning. If the overall assay QC fails, the assay receives an assay QC warning (across all samples). The NPX data are presented on a log2 scale, where an increase of one NPX unit corresponds to a doubling of the protein content. A high NPX value indicates a high protein concentration. Each measured protein has a LOD determined at run time based on negative controls. Values <LOD were retained in the analyses, unless stated otherwise, but considered missing values. Statistical analysis Protein overlaps and reference plasma proteome The reference plasma proteome was compiled from proteins listed in the Human Plasma section of the PeptideAtlas (build 2023-04, www.peptideatlas.org) ( 26 ), proteins with an estimated blood concentration in the HPA (v24, www.proteinatlas.org) ( 27 , 28 ), and proteins classified as secreted to blood in the HPA, resulting in a reference set of 4,889 unique proteins based on UniProt IDs. Some proteins in the HPA data lacked UniProt IDs. Where possible, these were assigned UniProt IDs from the MS and Olink data through gene name matching. Proteins overlapping between MS, Olink, PeptideAtlas, and HPA datasets were identified by exact UniProt ID matching. Estimated concentration in blood The estimated concentration in the blood of each protein was obtained from the HPA ( 28 ) and converted to ng/mL. The concentrations in the HPA are derived from immunoassay data from the literature and MS data from the PeptideAtlas ( 26 ). For proteins that had both immunoassay- and MS-based concentrations available, the MS-based concentration was used. For proteins lacking an MS-based concentration, the immunoassay-based concentration was used, if available. Comparison of protein annotations The frequency of specific HPA annotations was calculated among all proteins detected in at least one sample by MS or Olink. Overrepresentation of these annotations was tested using a hypergeometric test, with all proteins detected by MS and/or Olink used as the background (N = 4,362). P-values were adjusted for multiple testing using the FDR method, with a significance threshold of FDR < 0.05. The “Enriched tissue” category refers to proteins annotated as “Tissue enriched” in the Tissue section of the HPA ( 27 ), meaning their mRNA expression was at least four-fold higher in a specific tissue compared to all other tissues. ORA of GO Biological Processes between platforms was performed using the compareCluster function in the clusterProfiler (version 4.14.4) ( 46 ) R package. All proteins detected with MS and/or Olink were used as the background protein list (N = 4,362). Fold enrichment for GO terms was calculated as the ratio of the frequency of the input proteins (GeneRatio) to the frequency of the background proteins (BgRatio) found in the respective GO term gene list, based on UniProt IDs. For Olink assays with multiple UniProt IDs, the first one was used for analysis. P-values were adjusted for multiple testing using the FDR method, with a significance threshold of FDR < 0.05. The coverage of FDA-approved plasma protein biomarkers was based on a list compiled by Anderson ( 29 ). Proteins were matched between this list and the MS and Olink datasets based on UniProt IDs. Ten markers with no UniProt ID were excluded from the analysis. Technical coefficients of variation Technical CVs were calculated per protein and duplicate sample using the CV formula for data on a log 2 -scale ( 47 ): For proteins measured in multiple duplicate samples, the final technical CV was the mean of the individual duplicate CVs. Six samples were run in duplicate with MS, with each sample’s duplicates placed in different TMT sets. However, due to differences in detected proteins between TMT sets (missing values), each protein in the MS data had between zero and six duplicates available for the CV calculation. In the Olink data, technical CVs were calculated on values above the LOD of one duplicate of the Olink Sample Control, which is a standard pooled plasma sample. Because values 100%) were excluded from the analysis. This applied to four proteins in the MS data. Differential abundance analysis DAA was performed for overlapping proteins with no missing values (N = 569) between females (N = 37) and males (N = 51) using a two-sided Welch’s t-test, with males as the reference group. P-values were adjusted for multiple testing per platform using the FDR method, with a significance threshold of FDR < 0.05. The categorical agreement of the log2-fold change values was calculated as the percentage of proteins showing the same direction of change (positive or negative) in both platforms. Cross-platform correlation analyses Correlations between MS and Olink measurements of matched proteins were calculated using Spearman’s rank correlation coefficient, and all correlations were presented without filtering for statistical significance. Proteins with less than eight overlapping data points were excluded from the analyses. Correlations were calculated on a complete dataset of all overlapping proteins (N = 1129), a cleaned dataset where values <LOD and QC warnings in the Olink data were set to missing (N = 1064), and cleaned datasets additionally filtered for a maximum of 50% missing values (N = 791) or no missing values (N = 463). The cross-platform correlations were divided into categories of no correlation: r ∈ [-1, 0.3); weak correlation: r ∈ [0.3, 0.5); moderate correlation: r ∈ [0.5, 0.7); and strong correlation: r ∈ [0.7, 1.0]. Associations between the MS-Olink correlations and technical factors were assessed using both Spearman’s and Pearson’s correlation coefficients, at a significance level of α = 0.05. Information on protein mass, length, and number of isoforms were obtained from UniProt release 2023_03, and protein concentration from the HPA as described above. All other technical factors analyzed were obtained or calculated from the MS and Olink data files. For the peptide-level analysis, MS peptides were matched to Olink assays based on gene name ( i.e. Olink assay name), and correlations were calculated using Spearman’s rank correlation coefficient. Peptides quantified in less than 15 samples, and genes with less than two peptides were excluded from the analysis. Information on the sequence positions of different isoforms and cleavage products of AMBP, HYOU1, and MASP1 were obtained manually from UniProt (release 2025_01, https://www.uniprot.org/). The fasta sequences of proteins included in the PeptOlink R Shiny app were downloaded from UniProt (release 2022_05). Enrichment analyses GSEAs of GO terms, HPA annotations, and MSigDB, KEGG, and Reactome gene sets were performed using the clusterProfiler R package, with a ranked list of the MS-Olink correlations of all overlapping proteins as the input (N = 1129). P-values were adjusted for multiple testing using the FDR method, with a significance threshold of FDR < 0.05. ORAs of GO terms, HPA annotations, and MSigDB, KEGG, and Reactome gene sets among proteins in the low correlation (ρ < 0.3) and strong correlation (ρ ≥ 0.7) categories were performed as described for the comparison of GO Biological Processes between platforms. All overlapping proteins (N = 1129) were used as the background. Comparison with previous studies The previous studies included in the comparison of cross-platform correlations are summarized in Table S17. Cross-platform correlations were obtained from the supplementary materials of all publications except Petrera et al . ( 24 ), for which MS and Olink data files were downloaded, and correlations were calculated between the DDA-MS and Olink measurements of all overlapping proteins matched by UniProt IDs. Proteins were matched between studies primarily by UniProt IDs, or by gene name if UniProt IDs were not provided. Several of the studies comparing Olink and SomaScan calculated Olink-SomaScan correlations both on normalized and non-normalized SomaScan data. For these studies, the correlations calculated on the normalized data were used in the cross-study comparison. For the comparison of correlations by confidence tiers defined in Eldjarn et al. ( 15 ), data were obtained from Supplementary Table 29 of the original publication. The authors defined the confidence tiers as follows: tier 1, the highest confidence tier, included proteins with an Olink-SomaScan correlation >0.5 and cis -pQTLs detected on both platforms; tier 2 included proteins with a correlation of ≤0.5 and a cis -pQTL detected on both platforms; and tier 3 included proteins with a cis -pQTLs identified on only one or neither platform. Software All statistical analyses were performed in R (version 4.4.2) ( 48 ). Figures were assembled in Adobe Illustrator 2025 (version 29.3.1). Data visualization Data visualization was performed in R using the ggplot2, ggpubr, ggpp, ggExtra, ggrepel, patchwork, RColorBrewer, and colorspace packages. Venn diagrams were generated using the eulerr R package. Point densities were calculated and visualized using the ggpointdensity function in the ggpointdensity package. The geom_smooth function from the ggplot2 R package was used to add regression lines to scatter plots. The peptide correlation plots were created using the plotly and heatmaply R packages. Declarations Data availability The mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD061144. The affinity proteomics data have been deposited to the PRIDE repository with the dataset identifier PAD000006. All other data supporting the findings of this study are available within the paper and its supplementary information files. Code availability The R code used for the analyses is deposited at https://github.com/noorasissala/MS-Olink-comparison. The code for the PeptOlink R Shiny app is deposited at https://github.com/isabelle-leo/PeptOlink. Acknowledgements We acknowledge support from the Global Proteomics and Proteogenomics Unit and the Affinity Proteomics Unit at the Science for Life Laboratory, as well as the National Genomics Infrastructure Uppsala. The project was funded by the Science for Life Laboratory Technology Development Grant 2022 (MP). Author contributions Conceptualization: MP, MÅ, CF, NS. Data Curation: JF. Formal Analysis: NS, IL. Funding Acquisition: MP, MÅ, CF, JL, LEE. Methodology: MP, MÅ, CF, NS, IL, HB. Resources: MP, MÅ, CF, LEE, JL. Software: IL, NS. Investigation: XC, MÅ. Visualization: NS, IL. Supervision: MP, JL, HB, LEE. Writing—original draft: NS, MP. Writing—review & editing: NS, HB, IL, XC, JF, LEE, JL, CF, MÅ, MP. Competing interests The authors declare no competing interests. References B. He, Z. Huang, C. Huang, E. C. Nice, Clinical applications of plasma proteomics and peptidomics: Towards precision medicine. PROTEOMICS – Clin. Appl. 16 , 2100097 (2022). Y. Liu, A. Beyer, R. Aebersold, On the Dependency of Cellular Protein Levels on mRNA Abundance. Cell 165 , 535–550 (2016). P. E. Geyer, L. M. Holdt, D. Teupser, M. Mann, Revisiting biomarker discovery by plasma proteomics. Mol. Syst. Biol. 13 , 942 (2017). N. L. Anderson, N. G. Anderson, The Human Plasma Proteome: History, Character, and Diagnostic Prospects*. Mol. Cell. Proteomics 1 , 845–867 (2002). L. Dayon, O. Cominetti, M. Affolter, Proteomics of human biological fluids for biomarker discoveries: technical advances and recent applications. Expert Rev. Proteomics 19 , 131–151 (2022). J. M. Bader, V. Albrecht, M. Mann, MS-Based Proteomics of Body Fluids: The End of the Beginning. Mol. Cell. Proteomics 22 , 100577 (2023). M. Pernemalm, A. Sandberg, Y. Zhu, J. Boekel, D. Tamburro, J. M. Schwenk, A. Björk, M. Wahren-Herlenius, H. Åmark, C.-G. Östenson, M. Westgren, J. Lehtiö, In-depth human plasma proteome analysis captures tissue proteins and transfer of protein variants across the placenta. eLife 8 , e41608 (2019). H. Babačić, W. Christ, J. E. Araújo, G. Mermelekas, N. Sharma, J. Tynell, M. García, R. Varnaite, H. Asgeirsson, H. Glans, J. Lehtiö, S. Gredmark-Russ, J. Klingström, M. Pernemalm, Comprehensive proteomics and meta-analysis of COVID-19 host response. Nat. Commun. 14 , 5921 (2023). M. Tognetti, K. Sklodowski, S. Müller, D. Kamber, J. Muntel, R. Bruderer, L. Reiter, Biomarker Candidates for Tumors Identified from Deep-Profiled Plasma Stem Predominantly from the Low Abundant Area. J. Proteome Res. 21 , 1718–1735 (2022). K. Suhre, G. R. Venkataraman, H. Guturu, A. Halama, N. Stephan, G. Thareja, H. Sarwath, K. Motamedchaboki, M. K. R. Donovan, A. Siddiqui, S. Batzoglou, F. Schmidt, Nanoparticle enrichment mass-spectrometry proteomics identifies protein-altering variants for precise pQTL mapping. Nat. Commun. 15 , 989 (2024). K. Suhre, M. I. McCarthy, J. M. Schwenk, Genetics meets proteomics: perspectives for large population-based studies. Nat. Rev. Genet. 22 , 19–37 (2021). B. B. Sun, J. Chiou, M. Traylor, C. Benner, Y.-H. Hsu, T. G. Richardson, P. Surendran, A. Mahajan, C. Robins, S. G. Vasquez-Grinnell, L. Hou, E. M. Kvikstad, O. S. Burren, J. Davitte, K. L. Ferber, C. E. Gillies, Å. K. Hedman, S. Hu, T. Lin, R. Mikkilineni, R. K. Pendergrass, C. Pickering, B. Prins, D. Baird, C.-Y. Chen, L. D. Ward, A. M. Deaton, S. Welsh, C. M. Willis, N. Lehner, M. Arnold, M. A. Wörheide, K. Suhre, G. Kastenmüller, A. Sethi, M. Cule, A. Raj, H. M. Kang, L. Burkitt-Gray, E. Melamud, M. H. Black, E. B. Fauman, J. M. M. Howson, H. M. Kang, M. I. McCarthy, P. Nioi, S. Petrovski, R. A. Scott, E. N. Smith, S. Szalma, D. M. Waterworth, L. J. Mitnaul, J. D. Szustakowski, B. W. Gibson, M. R. Miller, C. D. Whelan, Plasma proteomic associations with genetics and health in the UK Biobank. Nature 622 , 329–338 (2023). A. M. Shah, P. L. Myhre, V. Arthur, P. Dorbala, H. Rasheed, L. F. Buckley, B. Claggett, G. Liu, J. Ma, N. Q. Nguyen, K. Matsushita, C. Ndumele, A. Tin, K. Hveem, C. Jonasson, H. Dalen, E. Boerwinkle, R. C. Hoogeveen, C. Ballantyne, J. Coresh, T. Omland, B. Yu, Large scale plasma proteomics identifies novel proteins and protein networks associated with heart failure development. Nat. Commun. 15 , 528 (2024). L. Wik, N. Nordberg, J. Broberg, J. Björkesten, E. Assarsson, S. Henriksson, I. Grundberg, E. Pettersson, C. Westerberg, E. Liljeroth, A. Falck, M. Lundberg, Proximity Extension Assay in Combination with Next-Generation Sequencing for High-throughput Proteome-wide Analysis. Mol. Cell. Proteomics 20 , 100168 (2021). G. H. Eldjarn, E. Ferkingstad, S. H. Lund, H. Helgason, O. T. Magnusson, K. Gunnarsdottir, T. A. Olafsdottir, B. V. Halldorsson, P. I. Olason, F. Zink, S. A. Gudjonsson, G. Sveinbjornsson, M. I. Magnusson, A. Helgason, A. Oddsson, G. H. Halldorsson, M. K. Magnusson, S. Saevarsdottir, T. Eiriksdottir, G. Masson, H. Stefansson, I. Jonsdottir, H. Holm, T. Rafnar, P. Melsted, J. Saemundsdottir, G. L. Norddahl, G. Thorleifsson, M. O. Ulfarsson, D. F. Gudbjartsson, U. Thorsteinsdottir, P. Sulem, K. Stefansson, Large-scale plasma proteomics comparisons through genetics and disease associations. Nature 622 , 348–358 (2023). D. H. Katz, J. M. Robbins, S. Deng, U. A. Tahir, A. G. Bick, A. Pampana, Z. Yu, D. Ngo, M. D. Benson, Z.-Z. Chen, D. E. Cruz, D. Shen, Y. Gao, C. Bouchard, M. A. Sarzynski, A. Correa, P. Natarajan, J. G. Wilson, R. E. Gerszten, Proteomic profiling platforms head to head: Leveraging genetics and clinical traits to compare aptamer- and antibody-based methods. Sci. Adv. 8 , eabm5164 (2022). M. R. Rooney, J. Chen, C. M. Ballantyne, R. C. Hoogeveen, O. Tang, M. E. Grams, A. Tin, C. E. Ndumele, F. Zannad, D. J. Couper, W. Tang, E. Selvin, J. Coresh, Comparison of Proteomic Measurements Across Platforms in the Atherosclerosis Risk in Communities (ARIC) Study. Clin. Chem. 69 , 68–79 (2023). L. M. Raffield, H. Dang, K. A. Pratte, S. Jacobson, L. A. Gillenwater, E. Ampleford, I. Barjaktarevic, P. Basta, C. B. Clish, A. P. Comellas, E. Cornell, J. L. Curtis, C. Doerschuk, P. Durda, C. Emson, C. M. Freeman, X. Guo, A. T. Hastie, G. A. Hawkins, J. Herrera, W. C. Johnson, W. W. Labaki, Y. Liu, B. Masters, M. Miller, V. E. Ortega, G. Papanicolaou, S. Peters, K. D. Taylor, S. S. Rich, J. I. Rotter, P. Auer, A. P. Reiner, R. P. Tracy, D. Ngo, R. E. Gerszten, W. K. O’Neal, R. P. Bowler, NHLBI Trans-Omics for Precision Medicine (TOPMed) Consortium, Comparison of Proteomic Assessment Methods in Multiple Cohort Studies. Proteomics 20 , e1900278 (2020). M. Pietzner, E. Wheeler, J. Carrasco-Zanini, A. Cortes, M. Koprulu, M. A. Wörheide, E. Oerton, J. Cook, I. D. Stewart, N. D. Kerrison, J. Luan, J. Raffler, M. Arnold, W. Arlt, S. O’Rahilly, G. Kastenmüller, E. R. Gamazon, A. D. Hingorani, R. A. Scott, N. J. Wareham, C. Langenberg, Mapping the proteo-genomic convergence of human diseases. Science 374 , eabj1541 (2021). F. Finkernagel, S. Reinartz, M. Schuldner, A. Malz, J. M. Jansen, U. Wagner, T. Worzfeld, J. Graumann, E. P. von Strandmann, R. Müller, Dual-platform affinity proteomics identifies links between the recurrence of ovarian carcinoma and proteins released into the tumor microenvironment. Theranostics 9 , 6601–6617 (2019). D. E. Haslam, J. Li, S. T. Dillon, X. Gu, Y. Cao, O. A. Zeleznik, N. Sasamoto, X. Zhang, A. H. Eliassen, L. Liang, M. J. Stampfer, S. Mora, Z.-Z. Chen, K. L. Terry, R. E. Gerszten, F. B. Hu, A. T. Chan, T. A. Libermann, S. N. Bhupathiraju, Stability and reproducibility of proteomic profiles in epidemiological studies: comparing the Olink and SOMAscan platforms. PROTEOMICS 22 , 2100170 (2022). E. B. Dammer, L. Ping, D. M. Duong, E. S. Modeste, N. T. Seyfried, J. J. Lah, A. I. Levey, E. C. B. Johnson, Multi-platform proteomic analysis of Alzheimer’s disease cerebrospinal fluid and plasma reveals network biomarkers associated with proteostasis and the matrisome. Alzheimers Res. Ther. 14 , 174 (2022). J. Graumann, F. Finkernagel, S. Reinartz, T. Stief, D. Brödje, H. Renz, J. M. Jansen, U. Wagner, T. Worzfeld, E. Pogge von Strandmann, R. Müller, Multi-platform Affinity Proteomics Identify Proteins Linked to Metastasis and Immune Suppression in Ovarian Cancer Plasma. Front. Oncol. 9 , 1150 (2019). A. Petrera, C. von Toerne, J. Behler, C. Huth, B. Thorand, A. Hilgendorff, S. M. Hauck, Multiplatform Approach for Plasma Proteomics: Complementarity of Olink Proximity Extension Assay Technology to Mass Spectrometry-Based Protein Profiling. J. Proteome Res. 20 , 751–762 (2021). A. K. Nilsson, H. Rydbeck, A. Thorsell, S. Frändberg, H. Barreto Henriksson, C. Hesse, G. Hellgren, P. Lundgren, A. Hellström, The proteome signature of cord blood plasma with high hematopoietic stem and progenitor cell count. Stem Cell Res. 61 , 102752 (2022). E. W. Deutsch, G. S. Omenn, Z. Sun, M. Maes, M. Pernemalm, K. K. Palaniappan, N. Letunica, Y. Vandenbrouck, V. Brun, S. Tao, X. Yu, P. E. Geyer, V. Ignjatovic, R. L. Moritz, J. M. Schwenk, Advances and Utility of the Human Plasma Proteome. J. Proteome Res. 20 , 5241–5263 (2021). M. Uhlén, L. Fagerberg, B. M. Hallström, C. Lindskog, P. Oksvold, A. Mardinoglu, Å. Sivertsson, C. Kampf, E. Sjöstedt, A. Asplund, I. Olsson, K. Edlund, E. Lundberg, S. Navani, C. A.-K. Szigyarto, J. Odeberg, D. Djureinovic, J. O. Takanen, S. Hober, T. Alm, P.-H. Edqvist, H. Berling, H. Tegel, J. Mulder, J. Rockberg, P. Nilsson, J. M. Schwenk, M. Hamsten, K. von Feilitzen, M. Forsberg, L. Persson, F. Johansson, M. Zwahlen, G. von Heijne, J. Nielsen, F. Pontén, Tissue-based map of the human proteome. Science 347 , 1260419 (2015). M. Uhlén, M. J. Karlsson, A. Hober, A.-S. Svensson, J. Scheffel, D. Kotol, W. Zhong, A. Tebani, L. Strandberg, F. Edfors, E. Sjöstedt, J. Mulder, A. Mardinoglu, A. Berling, S. Ekblad, M. Dannemeyer, S. Kanje, J. Rockberg, M. Lundqvist, M. Malm, A.-L. Volk, P. Nilsson, A. Månberg, T. Dodig-Crnkovic, E. Pin, M. Zwahlen, P. Oksvold, K. von Feilitzen, R. S. Häussler, M.-G. Hong, C. Lindskog, F. Ponten, B. Katona, J. Vuu, E. Lindström, J. Nielsen, J. Robinson, B. Ayoglu, D. Mahdessian, D. Sullivan, P. Thul, F. Danielsson, C. Stadler, E. Lundberg, G. Bergström, A. Gummesson, B. G. Voldborg, H. Tegel, S. Hober, B. Forsström, J. M. Schwenk, L. Fagerberg, Å. Sivertsson, The human secretome. Sci. Signal. 12 , eaaz0274 (2019). N. L. Anderson, The Clinical Plasma Proteome: A Survey of Clinical Assays for Proteins in Plasma and Serum. Clin. Chem. 56 , 177–185 (2010). L. R. Heil, E. Damoc, T. N. Arrey, A. Pashkova, E. Denisov, J. Petzoldt, A. C. Peterson, C. Hsu, B. C. Searle, N. Shulman, M. Riffle, B. Connolly, B. X. MacLean, P. M. Remes, M. W. Senko, H. I. Stewart, C. Hock, A. A. Makarov, D. Hermanson, V. Zabrouskov, C. C. Wu, M. J. MacCoss, Evaluating the Performance of the Astral Mass Analyzer for Quantitative Proteomics Using Data-Independent Acquisition. J. Proteome Res. 22 , 3290–3300 (2023). N. G. Hendricks, S. D. Bhosale, A. J. Keoseyan, J. Ortiz, A. Stotland, S. Seyedmohammad, C. D. L. Nguyen, J. T. Bui, A. Moradian, S. M. Mockus, J. E. Van Eyk, An Inflection Point in High-Throughput Proteomics with Orbitrap Astral: Analysis of Biofluids, Cells, and Tissues. J. Proteome Res. 23 , 4163–4169 (2024). M. M. Savitski, T. Mathieson, N. Zinn, G. Sweetman, C. Doce, I. Becher, F. Pachl, B. Kuster, M. Bantscheff, Measuring and Managing Ratio Compression for Accurate iTRAQ/TMT Quantification. J. Proteome Res. 12 , 3586–3598 (2013). M. L. Gardner, M. A. Freitas, Multiple Imputation Approaches Applied to the Missing Value Problem in Bottom-Up Proteomics. Int. J. Mol. Sci. 22 , 9650 (2021). C. Fredolini, S. Byström, L. Sanchez-Rivera, M. Ioannou, D. Tamburro, F. Pontén, R. M. Branca, P. Nilsson, J. Lehtiö, J. M. Schwenk, Systematic assessment of antibody selectivity in plasma based on a resource of enrichment profiles. Sci. Rep. 9 , 8324 (2019). A. Andersson, J. Remnestål, B. Nellgård, H. Vunk, D. Kotol, F. Edfors, M. Uhlén, J. M. Schwenk, L. L. Ilag, H. Zetterberg, K. Blennow, A. Månberg, P. Nilsson, C. Fredolini, Development of parallel reaction monitoring assays for cerebrospinal fluid proteins associated with Alzheimer’s disease. Clin. Chim. Acta 494 , 79–93 (2019). K. Sikorski, A. Mehta, M. Inngjerdingen, F. Thakor, S. Kling, T. Kalina, T. A. Nyman, M. E. Stensland, W. Zhou, G. A. de Souza, L. Holden, J. Stuchly, M. Templin, F. Lund-Johansen, A high-throughput pipeline for validation of antibodies. Nat. Methods 15 , 909–912 (2018). M. Uhlen, A. Bandrowski, S. Carr, A. Edwards, J. Ellenberg, E. Lundberg, D. L. Rimm, H. Rodriguez, T. Hiltke, M. Snyder, T. Yamamoto, A proposal for validation of antibodies. Nat. Methods 13 , 823–827 (2016). A. Levitsky, M. Pernemalm, B.-M. Bernhardson, J. Forshed, K. Kölbeck, M. Olin, R. Henriksson, J. Lehtiö, C. Tishelman, L. E. Eriksson, Early symptoms and sensations as predictors of lung cancer: a machine learning multivariate model. Sci. Rep. 9 , 16504 (2019). X. Cao, A. Sandberg, J. E. Araújo, F. Cvetkovski, E. Berglund, L. E. Eriksson, M. Pernemalm, Evaluation of Spin Columns for Human Plasma Depletion to Facilitate MS-Based Proteomics Analysis of Plasma. J. Proteome Res. 20 , 4610–4620 (2021). R. M. M. Branca, L. M. Orre, H. J. Johansson, V. Granholm, M. Huss, Å. Pérez-Bercoff, J. Forshed, L. Käll, J. Lehtiö, HiRIEF LC-MS enables deep proteome coverage and unbiased proteogenomics. Nat. Methods 11 , 59–62 (2014). M. C. Chambers, B. Maclean, R. Burke, D. Amodei, D. L. Ruderman, S. Neumann, L. Gatto, B. Fischer, B. Pratt, J. Egertson, K. Hoff, D. Kessner, N. Tasman, N. Shulman, B. Frewen, T. A. Baker, M.-Y. Brusniak, C. Paulse, D. Creasy, L. Flashner, K. Kani, C. Moulding, S. L. Seymour, L. M. Nuwaysir, B. Lefebvre, F. Kuhlmann, J. Roark, P. Rainer, S. Detlev, T. Hemenway, A. Huhmer, J. Langridge, B. Connolly, T. Chadick, K. Holly, J. Eckels, E. W. Deutsch, R. L. Moritz, J. E. Katz, D. B. Agus, M. MacCoss, D. L. Tabb, P. Mallick, A Cross-platform Toolkit for Mass Spectrometry and Proteomics. Nat. Biotechnol. 30 , 918–920 (2012). P. Di Tommaso, M. Chatzou, E. W. Floden, P. P. Barja, E. Palumbo, C. Notredame, Nextflow enables reproducible computational workflows. Nat. Biotechnol. 35 , 316–319 (2017). S. Kim, P. A. Pevzner, MS-GF+ makes progress towards a universal database search tool for proteomics. Nat. Commun. 5 , 5277 (2014). V. Granholm, S. Kim, J. C. F. Navarro, E. Sjölund, R. D. Smith, L. Käll, Fast and Accurate Database Searches with MS-GF+Percolator. J. Proteome Res. 13 , 890–897 (2014). H. L. Röst, T. Sachsenberg, S. Aiche, C. Bielow, H. Weisser, F. Aicheler, S. Andreotti, H.-C. Ehrlich, P. Gutenbrunner, E. Kenar, X. Liang, S. Nahnsen, L. Nilse, J. Pfeuffer, G. Rosenberger, M. Rurik, U. Schmitt, J. Veit, M. Walzer, D. Wojnar, W. E. Wolski, O. Schilling, J. S. Choudhary, L. Malmström, R. Aebersold, K. Reinert, O. Kohlbacher, OpenMS: a flexible open-source software platform for mass spectrometry data analysis. Nat. Methods 13 , 741–748 (2016). S. Xu, E. Hu, Y. Cai, Z. Xie, X. Luo, L. Zhan, W. Tang, Q. Wang, B. Liu, R. Wang, W. Xie, T. Wu, L. Xie, G. Yu, Using clusterProfiler to characterize multiomics data. Nat. Protoc. 19 , 3292–3320 (2024). J. A. Canchola, S. Tang, P. Hemyari, E. Paxinos, E. Marins, Correct use of percent coefficient of variation (%CV) formula for log-transformed data. MOJ Proteomics Bioinforma. 6 , 316–317 (2017). R Core Team, R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria, 2024; https://www.R-project.org/). Additional Declarations The authors declare no competing interests. Supplementary Files Sissalapreprintsupplementarymaterials.docx Supplementary figures and table legends Sissalapreprintsupplementarytables.xlsx Supplementary tables Cite Share Download PDF Status: Published Journal Publication published 03 Nov, 2025 Read the published version in Communications Chemistry → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6501601","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":449781924,"identity":"21f4eecd-809c-407e-b6d3-4b91a48b7873","order_by":0,"name":"Noora Sissala","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA10lEQVRIiWNgGAWjYBADHiBmPMBQkcBDkhaGAwxnSNACBgcY2xIIqzJnbz724QdDnYx8/+EHB37OS5Nh4D98AK8Wy55jyTN7GA7zGNxIMzjYuy2Hh0EiDb9VBjdyjIEeOcBjIMFgcIB3WwVQC48BQS2MfxjqeOT7j384+HcOUAv/+Q8EtTDzMADRgRyDw7wNQIcx5ODVwWBw5lgys4wByC85BYdljqXxsEmkEXDY8ebDjG8q6uyBDtv48E1Nsj0//+EH+K2BaERisxGhfhSMglEwCkYBAQAADJ5BO3k2xq4AAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0000-0758-8140","institution":"Karolinska Institutet","correspondingAuthor":true,"prefix":"","firstName":"Noora","middleName":"","lastName":"Sissala","suffix":""},{"id":449781925,"identity":"ba0ca9f6-6f0d-4c7a-bc51-e6832ffddcff","order_by":1,"name":"Haris Babačić","email":"","orcid":"https://orcid.org/0000-0003-0813-0005","institution":"Karolinska Institutet","correspondingAuthor":false,"prefix":"","firstName":"Haris","middleName":"","lastName":"Babačić","suffix":""},{"id":449781926,"identity":"babe8675-55a3-4f62-8592-e556f993b1b3","order_by":2,"name":"Isabelle R. Leo","email":"","orcid":"https://orcid.org/0000-0002-7627-6690","institution":"Karolinska Institutet","correspondingAuthor":false,"prefix":"","firstName":"Isabelle","middleName":"R.","lastName":"Leo","suffix":""},{"id":449781927,"identity":"02b1cfa8-f89b-42d5-a5fb-9dfe27d8b75f","order_by":3,"name":"Xiaofang Cao","email":"","orcid":"","institution":"Karolinska Institutet","correspondingAuthor":false,"prefix":"","firstName":"Xiaofang","middleName":"","lastName":"Cao","suffix":""},{"id":449781928,"identity":"97fad38a-4c60-4714-937c-309f51f51b5d","order_by":4,"name":"Jenny Forshed","email":"","orcid":"","institution":"Karolinska Institutet","correspondingAuthor":false,"prefix":"","firstName":"Jenny","middleName":"","lastName":"Forshed","suffix":""},{"id":449781929,"identity":"55d8eef0-ec72-4368-963e-b77a75148b52","order_by":5,"name":"Lars E. Eriksson","email":"","orcid":"https://orcid.org/0000-0001-5121-5325","institution":"Karolinska Institutet","correspondingAuthor":false,"prefix":"","firstName":"Lars","middleName":"E.","lastName":"Eriksson","suffix":""},{"id":449781930,"identity":"f8d28b96-7e86-4efa-abc9-905b1f460498","order_by":6,"name":"Janne Lehtiö","email":"","orcid":"https://orcid.org/0000-0002-8100-9562","institution":"Karolinska Institutet","correspondingAuthor":false,"prefix":"","firstName":"Janne","middleName":"","lastName":"Lehtiö","suffix":""},{"id":449781931,"identity":"3b8b850d-b06f-499d-ad4d-5cbbf2feff68","order_by":7,"name":"Claudia Fredolini","email":"","orcid":"https://orcid.org/0000-0002-7674-2014","institution":"KTH Royal Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Claudia","middleName":"","lastName":"Fredolini","suffix":""},{"id":449781932,"identity":"4a9d912a-49e2-43ea-aa02-06daaecace08","order_by":8,"name":"Mikael Åberg","email":"","orcid":"https://orcid.org/0000-0002-7858-8233","institution":"Uppsala University","correspondingAuthor":false,"prefix":"","firstName":"Mikael","middleName":"","lastName":"Åberg","suffix":""},{"id":449781933,"identity":"b79d69b7-38db-4a96-b7a7-e5a870e8ada7","order_by":9,"name":"Maria Pernemalm","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9ElEQVRIiWNgGAWjYJCCAwhmAYOcBDNpWgwYjInSggQMGBJnEFKj23724YEfDHfkzGc3H93ww2Bb+sx2BuYPH/BoMTuTbnCwh+GZscydY2k3ewxu585mZmCTxGeV2YE0hgM8DIcTZ0jkmN3gAWqZB9TCzINPy/lnDAf/MByunyGR/+3mH4Pb6XLMDMyf/+DTciON4TDQlgQJiRy220BbEqSBISaNz/tmN54xHJYxeGY4QyLN7LaMwW3Dmc2MbZI9eB2WxvzxTcUdeQmJ5Gc331Tclpc4f/jwhx/4rAEDgwPIPMYGghoYUBLAKBgFo2AUjAJ0AACDRU+9uu3fmwAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-4624-031X","institution":"Karolinska Institutet","correspondingAuthor":true,"prefix":"","firstName":"Maria","middleName":"","lastName":"Pernemalm","suffix":""}],"badges":[],"createdAt":"2025-04-22 07:47:42","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6501601/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6501601/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s42004-025-01753-2","type":"published","date":"2025-11-04T00:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":81999079,"identity":"46fccdb2-4ec1-4773-8226-49fd51287476","added_by":"auto","created_at":"2025-05-05 19:02:43","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":296199,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOverview of the study.\u0026nbsp;\u003c/strong\u003ePlasma samples:\u003cstrong\u003e \u003c/strong\u003eA total of\u003cstrong\u003e \u003c/strong\u003e114 pre-diagnostic plasma samples were collected from patients under investigation for suspected lung cancer. Proteomic analysis: The 114 samples were analyzed using MS-based proteomics (HiRIEF LC-MS/MS). Six samples were run in duplicate, in different TMT sets, resulting in a total of 120 samples. A subset of 88 samples were also analyzed using the Olink Explore 3072 PEA-based platform, along with one duplicate control sample. The duplicate samples were used to calculate technical CVs. Cohort characteristics: Age and sex distribution of the study population (N = 88). Method comparison: We evaluated proteome coverage, precision, statistical power, and quantitative agreement between HIRIEF LC-MS/MS and Olink Explore 3072 at both the protein and peptide levels. In addition, we developed a publicly available resource, the PeptOlink R Shiny app, for exploring peptide-level correlations between the two technologies. Created in BioRender. https://BioRender.com/i93h320.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/df413a699cebbf71ab86d178.png"},{"id":81999077,"identity":"706beeb4-e8e8-4b07-ae14-a1a96267a7dd","added_by":"auto","created_at":"2025-05-05 19:02:43","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":356881,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDetected proteins, missing values, and plasma proteome coverage\u003c/strong\u003e. \u003cstrong\u003e(A)\u003c/strong\u003e Venn diagram of proteins detected (in at least one sample) with HiRIEF LC-MS/MS and/or Olink Explore 3072, based on unique UniProt IDs. \u003cstrong\u003e(B) \u003c/strong\u003eNumber and percentage of Olink assays in each Olink Explore panel detected with both Olink and MS (= overlapping proteins). \u0026nbsp;\u003cstrong\u003e(C)\u003c/strong\u003e Detected proteins and missing values in the MS and Olink datasets. The values on the y-axis indicate the percentage of proteins in each dataset with a percentage of missing values within the intervals defined on the x-axis. The dotted line indicates the cumulative number of proteins with a proportion of missing values less than or equal to the upper bound of each interval. \u003cstrong\u003e(D) \u003c/strong\u003eVenn diagram comparing proteins detected by MS and Olink to proteins in the reference human plasma proteome, compiled from the PeptideAtlas and the HPA (see Methods). The bar plot shows the proportion of proteins in the reference plasma proteome detected with MS, Olink, or both methods. \u003cstrong\u003e(E)\u003c/strong\u003e Distribution of the estimated concentrations of all detected proteins (left) and proteins detected exclusively by MS or Olink (right). Concentration data were downloaded from the HPA (28). Medians and inter-quartile ranges are indicated with points and error bars. Differences in protein concentration between platforms were tested using a Wilcoxon rank-sum test. \u003cstrong\u003e(F) \u003c/strong\u003eMissing values per protein by estimated protein concentration. Differences in protein concentration between platforms were tested using a Wilcoxon rank-sum test, and p-values were adjusted with the FDR method. ns = non-significant, ** = p \u0026lt; 0.01, **** = p \u0026lt; 0.001. \u003cstrong\u003e(G) \u003c/strong\u003ePlasma proteome coverage by estimated protein concentration. Each bar shows the proportion of proteins in the reference plasma proteome, within a specific concentration interval, that were detected with either MS only, Olink only, both methods, or neither method (“Not detected”). The x-axis intervals are right closed.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/e52b15fd0560dc7eddf3ab17.png"},{"id":81999080,"identity":"2a13b54d-b575-4f98-9e45-8956f266ab82","added_by":"auto","created_at":"2025-05-05 19:02:43","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":622774,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCharacterization of detected proteins.\u003c/strong\u003e \u003cstrong\u003e(A)\u003c/strong\u003eComparison of the frequency of protein annotations from select HPA categories among proteins detected with HiRIEF LC-MS/MS and Olink Explore 3072. The 15 most frequent annotations (for both platforms) within each HPA category are shown. The frequency was calculated by platform in relation to the total number of proteins detected by that platform that were also found in the HPA. Asterisks indicate statistically significant differences in frequency between platforms (Fisher test, FDR \u0026lt; 0.05). \u003cstrong\u003e(B)\u003c/strong\u003e ORA of GO Biological Processes among proteins detected with each technology. All proteins detected by MS and/or Olink were used as the background protein list (N = 4,362). Points are colored by the number of proteins from each GO term that were found in the target protein lists, that is, among the MS or Olink proteins.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/c6b751d72e06d9e4124185ae.png"},{"id":81999086,"identity":"b366be33-fcbf-45b4-abc2-5731472c951d","added_by":"auto","created_at":"2025-05-05 19:02:43","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":195414,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrecision and statistical power in DAA\u003c/strong\u003e.\u003cstrong\u003e (A) \u003c/strong\u003eTechnical CVs per protein for HiRIEF LC-MS/MS and Olink Explore 3072. Medians and inter-quartile ranges are indicated with points and error bars. \u003cstrong\u003e(B)\u003c/strong\u003e Number of DAPs between males and females by platform (Welch’s t-test, FDR \u0026lt; 0.05). The analysis was performed on overlapping proteins with no missing values to ensure equal sample sizes across platforms. \u003cstrong\u003e(C) \u003c/strong\u003e\u0026nbsp;Agreement of log2-fold change values (calculated as female – male) between platforms for all proteins included in the DAA (left) and for proteins that were DAPs in either platform (Welch’s t-test, FDR \u0026lt; 0.05) (right). Points are colored by which dataset(s) the protein was found to be differentially abundant in: MS only, Olink only, both, or none. The solid black line represents perfect agreement between the platforms, with a slope of 1. The categorical agreement of the log2-fold change values was calculated as the percentage of proteins showing the same direction of change (positive or negative) in both platforms. \u003cstrong\u003e(D) \u003c/strong\u003eEstimated concentration in blood from the HPA for DAPs found by MS only, Olink only, or both platforms (Welch’s t-test, FDR \u0026lt; 0.05). Differences between categories were tested using the Wilcoxon rank-sum test, and p-values were adjusted for multiple testing with the FDR method. ns = not significant, * = p \u0026lt; 0.05, *** = p \u0026lt; 0.001. The density plot on the right shows the distribution of estimated protein concentration by platform for all DAPs.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/84af2c9f64988eb188541c15.png"},{"id":81999586,"identity":"becc82ce-d7a6-4c54-a246-ebbeebf4592a","added_by":"auto","created_at":"2025-05-05 19:10:43","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":525646,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-platform correlation of protein levels. (A) \u003c/strong\u003eHistogram of Spearman’s rank correlations between HiRIEF LC-MS/MS and Olink Explore 3072 protein measurements for overlapping proteins (N = 1129). Correlations were categorized into groups of no correlation: r ∈ [-1, 0.3); weak correlation: r ∈ [0.3, 0.5); moderate correlation: r ∈ [0.5, 0.7); and strong correlation: r ∈ [0.7, 1.0]. The dashed line indicates the median correlation.\u003cstrong\u003e (B)\u003c/strong\u003e Proteins ranked by MS-Olink correlation and colored by proportion of samples with a missing value in either MS or Olink. \u003cstrong\u003e(C)\u003c/strong\u003e Same as A, but for overlapping proteins with no missing values (N = 463). \u003cstrong\u003e(D) \u003c/strong\u003eScatter plot of the MS-Olink correlations from the current study vs. the Olink-SomaScan correlations for the same proteins from Eldjarn et al. (15)(matched by UniProt ID). Points are colored by the confidence tier assigned by Eldjarn et al.; tier 1 includes proteins with an Olink-SomaScan correlation above 0.5 and a cis-pQTL in both platforms, tier 2 includes proteins with a correlation of 0.5 or lower and a cis-pQTL in both platforms, and tier 3 includes proteins with a cis-pQTL in only one or none of the platforms. The dashed line represents perfect agreement between studies, with a slope of 1. The density plots around the plot margins show the distribution of correlations by confidence tier, with dashed lines indicating the median in each tier.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/ad7d2a1e3809f68fa229663d.png"},{"id":81999591,"identity":"bbf649b0-4d3d-479e-bd83-3116111ed7a0","added_by":"auto","created_at":"2025-05-05 19:10:43","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":791972,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEffect of data quality factors and protein properties on cross-platform correlations. (A) \u003c/strong\u003eSpearman correlation between HiRIEF LC-MS/MS and Olink Explore 3072 proteins by the proportion of missing values per protein in MS. Proteins with any missing value in Olink data were excluded from analysis. \u003cstrong\u003e(B) \u003c/strong\u003eMS-Olink correlations by the proportion of missing values per protein in Olink. Proteins with any missing value in MS data were excluded from analysis. Differences in correlation between intervals were tested, in both A and B, using the Wilcoxon rank-sum test, and p-values were adjusted for multiple testing with the FDR method. ns = not significant, * = p \u0026lt; 0.05, ** = p \u0026lt; 0.01. \u003cstrong\u003e(C)\u003c/strong\u003e MS-Olink correlation vs. median number of PSMs used for quantification across TMT sets with MS. \u003cstrong\u003e(D)\u003c/strong\u003e MS-Olink correlation vs. median number of unique peptides used for quantification across TMT sets with MS. The x-axes in C and D are on a log\u003csub\u003e10\u003c/sub\u003e-scale.­\u003cstrong\u003e (E)\u003c/strong\u003e Left: Comparison of the MS-Olink correlation of NPX values with (“Warn”) and without (“Pass”) a sample QC warning in the Olink data, calculated per sample. Right: ­Comparison of the MS-Olink correlation of proteins with (“Warn”) and without (“Pass”) an assay QC warning in the Olink data. Statistical significance was determined as in A-B. ns = not significant, **** = p \u0026lt; 0.0001. \u003cstrong\u003e(F)\u003c/strong\u003e HPA annotations enriched among MS-Olink correlations at 5% FDR. Each row shoes the distribution of MS-Olink correlations for proteins associated with a specific HPA annotation.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/9cea86221b8dbcd8773bf377.png"},{"id":81999089,"identity":"dfee1e72-aa07-4baf-a008-3e3d938a391f","added_by":"auto","created_at":"2025-05-05 19:02:43","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":609437,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-platform correlations at the peptide level for AMBP, HYOU1, and MASP1. \u003c/strong\u003eVisualization of cross-platform correlations between HiRIEF LC-MS/MS peptides and Olink Explore 3072 assays for three representative proteins: \u003cstrong\u003e(A) \u003c/strong\u003eAMBP, \u003cstrong\u003e(B) \u003c/strong\u003eHYOU1, and \u003cstrong\u003e(C) \u003c/strong\u003eMASP1 (isoform 2). For each protein, MS peptides and select sequence features (y-axis) are plotted along the protein sequence (x-axis). Each horizontal bar represents either a sequence feature (middle part of each plot) or a peptide (bottom part of each plot). Peptides are shown in multiple rows to avoid overlaps and are colored by their correlation to the corresponding Olink assay. At the top of each plot, a density graph illustrates the distribution of peptide-Olink correlations across the protein sequence, divided into categories of no correlation: r ∈ [-1, 0.3); weak correlation: r ∈ [0.3, 0.5); moderate correlation: r ∈ [0.5, 0.7); and strong correlation: r ∈ [0.7, 1.0]. The color legends for the correlation category and the MS-Olink correlation (shown in A) apply to all subfigures.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/25db7de7b2054f29a9b3168b.png"},{"id":101163511,"identity":"0097b8d6-94ba-44cd-88cd-c22331266d6b","added_by":"auto","created_at":"2026-01-26 19:42:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4553166,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/59bad71b-b2e2-4b4a-9d8e-2b5a6925fd33.pdf"},{"id":81999928,"identity":"80a54e3d-529a-45d6-ba3e-80b86d20ac61","added_by":"auto","created_at":"2025-05-05 19:18:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4415488,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary figures and table legends\u003c/p\u003e","description":"","filename":"Sissalapreprintsupplementarymaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/eaef1629ceb6ebd70015118e.docx"},{"id":81999081,"identity":"62334780-8183-41f1-a031-2c9b515aa06f","added_by":"auto","created_at":"2025-05-05 19:02:43","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":4443875,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary tables\u003c/p\u003e","description":"","filename":"Sissalapreprintsupplementarytables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6501601/v1/42dc0fcb92964bf490580280.xlsx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eComparative evaluation of in-depth mass spectrometry and Olink Explore 3072 for plasma proteome profiling\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePrecision medicine and precision health depend on biomarkers to guide and facilitate disease prevention, diagnosis, prognosis, and treatment in an increasingly personalized manner. Blood plasma is a promising source of potential biomarkers, as it can be sampled in a minimally invasive manner and contains valuable biological and physiological information from all the organs in the body, harbored in a complex mixture of molecules such as nucleic acids, lipids, metabolites, and proteins (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Proteins can be especially useful as biomarkers, since their levels are closely associated with the phenotype of the individual. Proteins are the effector molecules of cells and tissues, and their levels vary in different physiological and pathological states in a way that cannot fully be predicted by the genome or transcriptome (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). The significance of proteins for clinical practice is demonstrated by the fact that around 40% of laboratory blood tests in the clinic measure proteins (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWhile the plasma proteome holds great potential for biomarker discovery, analyzing it has proven challenging due to its high dynamic range of concentrations. Protein concentrations in plasma span at least 12 orders of magnitude. Among the thousands of proteins in plasma, the 22 most abundant constitute 99% of the plasma proteome by mass (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). Disease-related proteins are often present at low levels in plasma and their detection has historically required either extensive sample processing for untargeted analysis or targeted analysis of individual or small sets of proteins. However, recent advancements in both global mass spectrometry (MS) and highly multiplexed affinity-based proteomics techniques have alleviated this problem by simultaneously increasing proteome coverage and sample throughput (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Consequently, larger cohorts can be profiled comprehensively, increasing the potential for insights into human health and disease and facilitating the discovery of new biomarkers.\u003c/p\u003e \u003cp\u003eIn global MS-based approaches, proteins are measured in an untargeted manner by digesting proteins into peptides, separating and ionizing the peptides, measuring their mass-to-charge ratios with MS, and identifying and quantifying the peptides by matching their mass spectra to theoretical mass spectra from sequence databases (peptide-spectrum matching). Generally, MS-based approaches offer highly specific identification and quantification of detected proteins (peptides) but require time-consuming sample preparation to profile the plasma proteome in depth. Thus, while MS-based approaches can reach an analytical depth of thousands of proteins, studies employing MS proteomics have been limited in their sample size compared to those employing affinity-based methods (\u003cspan additionalcitationids=\"CR8 CR9\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn contrast, affinity-based approaches, where affinity molecules such as antibodies or aptamers are used to bind and quantify predefined target proteins, enable high-throughput profiling of the plasma proteome (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). These methods, including the antibody-based proximity extension assays (PEAs) from Olink and the aptamer-based SomaScan assays from SomaLogic, have facilitated large-scale studies involving thousands of individuals (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). However, unlike MS, these methods do not provide direct detection of proteins (peptides), and ensuring the specificity and accuracy of the affinity binders is a challenge. To help mitigate this issue, PEAs rely on two antibodies to detect each target protein (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eUnderstanding the relative strengths and limitations of the various technologies available for plasma proteome profiling is essential to allow informed decisions regarding study design, platform selection, and data quality control (QC). Numerous studies have compared the performance of Olink\u0026rsquo;s PEAs and SomaLogic\u0026rsquo;s SomaScan assays (\u003cspan additionalcitationids=\"CR16 CR17 CR18 CR19 CR20 CR21 CR22\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e), but few have compared MS and Olink\u0026rsquo;s PEAs, and these comparisons have been limited in their profiling depth (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Here we present a comprehensive comparative evaluation of the Olink Explore 3072 PEA-based platform and our previously published method for in-depth MS-based plasma proteomics, high-resolution isoelectric focusing coupled with liquid chromatography and tandem mass spectrometry (HiRIEF LC-MS/MS) (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). This workflow involves depletion of high-abundance proteins, tandem mass tag (TMT) labeling, extensive pre-fractionation of peptides using HiRIEF, and data-dependent acquisition (DDA) to achieve high analytical depth and relative quantification. We evaluate the two methods in terms of proteome coverage, precision, statistical power, and quantitative agreement at both the protein and peptide level. Finally, we present PeptOlink, a public resource for analyzing the agreement between peptide and protein abundances, as quantified by MS and Olink, respectively.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eOverview of the proteomics datasets\u003c/p\u003e\n\u003cp\u003eWe detected 2,578 unique proteins across 120 samples using HiRIEF LC-MS/MS (114 distinct samples with six samples run in duplicate) and measured 2,923 proteins in a subset of 88 samples using Olink Explore 3072 (Fig.\u0026nbsp;1). In the Olink data, Normalized Protein Expression (NPX) values below the limit of detection (LOD) were retained but considered missing values. Ten proteins with NPX values below the LOD in all samples were deemed not detected and were excluded from further analysis. In total, 4,362 proteins were detected and quantified in at least one sample across both technologies, 2,578 with MS and 2,913 with Olink, with 1,129 overlapping between methods (Fig.\u0026nbsp;2A, Table S1). The number of overlapping proteins varied by Olink Explore panel, with the greatest overlap observed for the Cardiometabolic panel (Fig.\u0026nbsp;2B). While both technologies detected a large number of proteins, there was a clear difference in the frequency of missing values. In the MS data, 55% of all quantified proteins had at least one missing value, compared to 20% of proteins in the Olink data (Fig.\u0026nbsp;2C). A total of 1,822 proteins were detected in at least 50% of the samples with MS, and 2,460 with Olink, while 1,212 and 1,910 were detected in all samples with MS and Olink, respectively. In the MS data, missing values were TMT set-specific (Figure S1).\u003c/p\u003e\n\u003cp\u003ePlasma proteome coverage\u003c/p\u003e\n\u003cp\u003eTo compare the plasma proteome coverage of each method, we curated a reference set of 4,889 plasma proteins by compiling proteins from the PeptideAtlas (build 2023-04, www.peptideatlas.org) (26) and the Human Protein Atlas (HPA, www.proteinatlas.org) (27, 28) (see Methods). HiRIEF LC-MS/MS showed a greater overlap with this reference plasma proteome, while Olink Explore 3072 targeted more than a thousand proteins not reported in the MS-based studies found in the PeptideAtlas (Fig.\u0026nbsp;2D, Table S1). Combined, the platforms covered 63% of proteins included in the reference plasma proteome.\u003c/p\u003e\n\u003cp\u003eNext, we evaluated plasma proteome coverage across concentration ranges, based on the estimated concentration in blood of each protein from the HPA (28). Both technologies detected proteins with estimated plasma concentrations spanning 10 orders of magnitude, down to the pg/mL concentration range (Fig.\u0026nbsp;2E). However, low-abundance proteins frequently had a large proportion of missing values, especially in the MS data (Fig.\u0026nbsp;2F). Olink demonstrated a higher coverage of low-abundance proteins, while MS demonstrated a higher coverage of mid to high-abundance proteins (Fig.\u0026nbsp;2G). Therefore, proteins detected exclusively by Olink were mainly low abundance, whereas those detected exclusively by MS tended to have higher concentrations (Fig.\u0026nbsp;2E). These observations remained consistent when considering proteins detected in at least 50% of samples, although the coverage of low-abundance proteins decreased for both methods (Figure S2).\u003c/p\u003e\n\u003cp\u003eCharacterization of detected proteins\u003c/p\u003e\n\u003cp\u003eTo further characterize the types of proteins that could be measured with each technology, we compared the frequency of various annotations from the HPA, performed overrepresentation analyses (ORA) of gene ontology (GO) terms, and investigated the coverage of plasma protein biomarkers approved by the United States Food and Drug Administration (FDA). In this cohort, predicted secreted proteins, enzymes, metabolic proteins, immunoglobulins, proteins enriched in liver tissue, and potential drug targets were more frequent in the MS data, while predicted membrane proteins, CD markers, proteins secreted in the male reproductive system, and proteins enriched in the brain and testis were more frequent in the Olink data (Fig.\u0026nbsp;3A, Table S2). It should be noted that 95 proteins (3.7%) detected by MS were not found in the HPA, compared to only 22 proteins (0.76%) detected by Olink, which could slightly bias the comparison of annotation frequencies between platforms in favor of Olink. Consistent with the observation that MS detected proportionally more mid to high-abundance plasma proteins, MS was enriched for GO biological processes inherent to the plasma proteome—hemostasis, blood coagulation, complement activation, and metabolism (Fig.\u0026nbsp;3B, Table S3). In contrast, Olink was enriched for GO biological processes related to low-abundance signaling proteins, particularly cytokines. The two methods detected comparable numbers of FDA-approved plasma protein biomarkers (29)—74 (MS) and 72 (Olink) out of 99, with 55 biomarkers detected by both (Table S4). Biomarkers exclusively detected by MS included various transport and metabolic proteins, whereas Olink exclusively covered several hormones.\u003c/p\u003e\n\u003cp\u003ePrecision\u003c/p\u003e\n\u003cp\u003eTo evaluate the precision of repeated measurements with HiRIEF LC-MS/MS and Olink Explore 3072, we calculated technical coefficients of variation (CV) for each protein across duplicate samples (Table S5). For Olink, intra-assay CVs were computed for 1,797 proteins (62%) after excluding values \u0026lt; LOD, based on a control sample of pooled donor plasma run in duplicate on the same plate. For MS, inter-assay CVs were calculated using duplicates of patient samples, with each replicate run in a different TMT set. Due to variations in protein identifications between TMT sets (\u003cem\u003ei.e.\u003c/em\u003e, missing values), CVs could be calculated for 1,952 proteins (76%).\u003c/p\u003e\n\u003cp\u003eBoth MS and Olink demonstrated high precision, as indicated by low technical CVs (Fig.\u0026nbsp;4A). The median CV was slightly lower for Olink than for MS (5.7% vs. 6.8%, p = 1.7 × 10⁻¹⁶). Most proteins had a CV below 15% in both datasets—85% in the MS data and 80% in the Olink data—although Olink had a higher proportion of proteins with very low CVs, below 5% (40% vs. 35%). However, it is important to note that the Olink CVs were intra-assay CVs, while the MS CVs were inter-assay CVs. Therefore, the Olink CVs may have been underestimated. Technical CVs were higher for proteins with more missing values, as well as for proteins with lower estimated concentrations in the blood (Figure S3).\u003c/p\u003e\n\u003cp\u003eStatistical power\u003c/p\u003e\n\u003cp\u003eNext, we investigated how the precision of HiRIEF LC-MS/MS and Olink Explore 3072 affected the statistical power of the differential abundance analysis (DAA) between males and females in both datasets. To ensure an equal number of tests for both technologies and equal sample sizes in each test, we focused the analysis on proteins quantified with both methods with no missing values (N = 569). After correcting for multiple hypothesis testing, there were 82 (14%) and 118 (21%) differentially abundant proteins (DAPs) between the sexes in the MS and Olink data, respectively (Table S6). Of these, 53 (9%) were statistically significant in both datasets (Fig.\u0026nbsp;4B). Therefore, 65% of DAPs found with MS were also found with Olink, while only 45% of DAPs found with Olink were also found with MS. These observations suggest that Olink may have had higher statistical power in detecting DAPs, whereas DAPs identified with MS were more likely to be reproduced with Olink. The platforms agreed on the direction of the difference for most proteins (80%), regardless of statistical significance. When considering proteins significant in at least one platform, the agreement was close to perfect (95%) (Fig.\u0026nbsp;4C). Thus, although the two methods had strong concordance in estimating differences in protein levels between groups, only a portion of the differences were consistently statistically significant.\u003c/p\u003e\n\u003cp\u003eFinally, we investigated the effect of technical and biological variance on the statistical significance of sex-related differences in protein levels. The technical CVs in MS were slightly higher for DAPs identified only with Olink (Figure S4A), suggesting that technical noise in MS may have masked some protein level differences. Overall, the log2-fold change estimates of DAPs had higher absolute values in the Olink data (Figure S4B). DAPs identified exclusively by Olink had lower estimated concentrations in the blood than the rest of the DAPs (Fig.\u0026nbsp;4D). These findings suggest that Olink may have quantified some low-abundance proteins with higher precision, contributing to a larger number of DAPs. On the other hand, there were still many proteins, generally higher in abundance, that were detected as DAPs only with MS.\u003c/p\u003e\n\u003cp\u003eCross-platform correlation of protein levels\u003c/p\u003e\n\u003cp\u003eTo investigate the accuracy of protein identifications and quantifications, we estimated the quantitative agreement between HiRIEF LC-MS/MS and Olink Explore 3072 for each protein using Spearman’s rank correlation coefficient. On the complete dataset of overlapping proteins (N = 1129), the median correlation between paired MS and Olink protein measurements was ρ = 0.59, with nearly two thirds exhibiting a moderate to strong correlation of ρ ∈ [0.5,1.0] (Fig.\u0026nbsp;5A, Table S7). Several proteins had near perfect agreement, with the highest correlations observed for MBL2, PZP, ANGPT1, MYL3, DPT, and SHMT1 (ρ ∈ [0.95,0.97]).\u003c/p\u003e\n\u003cp\u003eIn contrast, some proteins had strong disagreement, with negative correlation, for example PAXX, SRPK2, GLIPR1, IL10RB, ISM2, and LSM1 (ρ ∈ [-0.71, -0.40]). Proteins with very low cross-platform correlations generally had many missing values (Fig.\u0026nbsp;5B). The correlations improved slightly after removing values \u0026lt; LOD and QC warnings in the Olink data (N = 1064), and further after additional filtering for proteins with a maximum of 50% missing values or QC warnings (N = 791) (Figure S5, Table S7). On the subset of overlapping proteins with no missing values or QC warnings (N = 463), the median correlation reached ρ = 0.68, with 81% of proteins having a moderate to high correlation between platforms (Fig.\u0026nbsp;5C, Table S7). Among proteins with no missing values, the lowest correlations were found for PPP1R9B, GDF2, SEMA3G, CTSL, ADGRF5, DAG1, and AGT (ρ ∈ [-0.23, -0.05]).\u003c/p\u003e\n\u003cp\u003eComparison with previous studies\u003c/p\u003e\n\u003cp\u003eNext, we compared the correlations between HiRIEF LC-MS/MS and Olink Explore 3072 measurements with those from previous studies that assessed the agreement between MS, Olink, and/or SomaScan platforms (see Methods). Most previous studies have focused on the agreement between Olink and SomaScan (15–22). On average, our estimates of MS-Olink correlations were higher than the Olink-SomaScan correlations reported in most of these studies (Figure S6, Table S8).\u003c/p\u003e\n\u003cp\u003eIn a recent large-scale comparison of Olink and SomaScan platforms, Eldjarn \u003cem\u003eet al\u003c/em\u003e. (15) categorized proteins into three confidence tiers according to the reliability of their measurements, with tier 1 representing highest confidence, and tier 3 lowest confidence. The classification was based on cross-platform correlations and the presence of protein quantitative trait loci (pQTLs) (see Methods). To provide additional orthogonal validation with MS, we examined our estimates of MS-Olink correlations in the context of these confidence tiers. Tier 1 proteins had a clearly higher median correlation between HiRIEF LC-MS/MS and Olink Explore 3072 (ρ = 0.72), compared to tier 2 (ρ = 0.53) and tier 3 (ρ = 0.47) proteins (Fig.\u0026nbsp;5D, Table S9). This supports the idea that the presence of pQTLs on both platforms, along with a high cross-platform correlation, is indicative of more accurate protein quantification. The median correlation between HiRIEF LC-MS/MS and Olink Explore 3072 for all confidence tiers were higher than the corresponding Olink-SomaScan correlations (Figure S7). However, the difference was most pronounced for tier 3 proteins, where the median HiRIEF-Olink correlation was ρ = 0.47, compared to ρ = 0.05 for Olink-SomaScan. These results suggest that the Olink assays for tier 3 proteins may be more accurate than previously indicated. Overall, our data provide orthogonal validation for the quantification accuracy of many Olink, and by extension SomaScan assays, in tier 1, and for a few assays in tier 3, based on strong cross-platform correlations in both studies (ρ \u0026gt; 0.7, Table S9).\u003c/p\u003e\n\u003cp\u003eTo date, only a few studies have reported correlations between MS and Olink measurements, with a limited number of overlapping proteins (22, 24, 25). These studies have shown varying levels of agreement between MS and Olink, with median correlations ranging from ρ = 0.27 to ρ = 0.56 for proteins overlapping with the present study (Figure S8, Table S10). On average, our MS-Olink correlations were higher than those of previous studies for the corresponding proteins. One of these studies, by Dammer \u003cem\u003eet al.\u003c/em\u003e (22), also assessed the agreement between MS and SomaScan measurements. Our MS-Olink correlations were comparable to the MS-SomaScan correlations reported in their analysis (Figure S9, Table S11).\u003c/p\u003e\n\u003cp\u003eTechnical factors affecting cross-platform correlations\u003c/p\u003e\n\u003cp\u003eTo understand the poor quantitative agreement observed for some proteins, we investigated the associations between the MS-Olink correlations and various data quality indicators (Table S6). First, we found that lower correlations were linked to more missing values in both platforms (Fig.\u0026nbsp;6A-B, Figure S10). The proportion of missing values had a very minor and inconsistent effect on the spread of the data, making it unlikely that the association between the missing values and the cross-platform correlations were driven by the spread of the data itself (Figure S11). The higher percentage of missing values likely indicated noisier quantification, as proteins with more missing values had lower estimated concentrations, higher technical CVs, and median values closer to the LOD in the Olink data (Figure S12). Consequently, all these factors were associated with weaker cross-platform correlations (Figure S13).\u003c/p\u003e\n\u003cp\u003eSecond, cross-platform correlations were positively associated with the number of peptide spectrum matches (PSMs) and peptides used for quantification by MS, as well as with sequence coverage (Fig.\u0026nbsp;6C-D, FigureS13). Proteins with fewer PSMs, peptides, and lower sequence coverage also tended to have more missing values, further explaining the observed relationship between missing values and weaker correlations (Figure S14). The correlations were lower for proteins with only one median PSM or peptide, but still moderate on average (median ρ = 0.42, Figure S15), suggesting that a low number of PSMs or peptides alone is not sufficient to deem a protein quantification unreliable. Additionally, proteins with high precursor mass errors (\u0026gt; 5) in the MS data had weak correlations with Olink (Figure S16A). Most, but not all, of these proteins also had a large proportion of missing values in MS (Figure S16B). In summary, while missing values seem to be the strongest data quality indicator for MS, factors such as PSMs, peptides, and precursor mass errors can provide additional insights and further explain variations in MS-Olink correlations.\u003c/p\u003e\n\u003cp\u003eFinally, we investigated how correlations were affected by QC warnings and differed between the Explore panels in the Olink data. Although MS-Olink correlations were lower among values with a sample QC warning, they only affected a small proportion of the proteins quantified in each sample (Figs.\u0026nbsp;6E and S17). Only 30 proteins that overlapped with MS had an assay QC warning and showed no difference in cross-platform correlation compared to the rest (Fig.\u0026nbsp;6E). The cross-platform correlation varied by Olink panel, likely driven by missing values, since panels with lower median correlation had a higher average proportion of missing values per protein (Figure S18A-B). Similarly, compared to version I panel proteins, correlations were somewhat lower for the version II panel proteins, which also had more missing values on average (Figure S18C-D). Thus, the impact of Olink data quality on the cross-platform correlations was likely largely driven by protein detectability, with a smaller effect from sample QC warnings.\u003c/p\u003e\n\u003cp\u003eProtein properties affecting cross-platform correlations\u003c/p\u003e\n\u003cp\u003eNext, we explored potential similarities in protein properties among proteins with poor correlations between MS and Olink. We found no difference in cross-platform correlations based on protein mass, length, or the number of isoforms reported in the UniProt database (Figure S19), and no enriched GO, MSigDB, KEGG, or Reactome gene sets by Gene Set Enrichment Analysis (GSEA) or ORA. However, enzymes, enzyme inhibitors, predicted secreted proteins, proteins secreted to the digestive system and the blood, as well as candidate cardiovascular disease genes from the HPA were enriched among proteins with high cross-platform correlations (Fig.\u0026nbsp;6F, Table S12). Most of these annotations were also overrepresented in the strong correlation group (ρ ∈ [0.7,1.0]) by ORA, while the no correlation group (ρ ∈ [-1,0.3)) had an overrepresentation of proteins related to intermediate filaments, mainly keratins (Table S13). These findings could be explained by the higher abundance of the proteins with strong MS-Olink correlations, while keratins could reflect sample contamination.\u003c/p\u003e\n\u003cp\u003eCross-platform correlations at the peptide level\u003c/p\u003e\n\u003cp\u003eLastly, we examined cross-platform correlations between HiRIEF LC-MS/MS and Olink Explore 3072 at the peptide level, to identify protein sequences or regions with varying correlations between platforms. We calculated the correlation between protein measurements of Olink assays and corresponding peptide measurements in the MS data, matched by gene name. To obtain more robust correlation estimates, we excluded peptides quantified in less than 15 samples with MS, resulting in a dataset of 13,856 peptides mapping to 822 genes and 847 unique UniProt IDs (Table S14). To enhance the accessibility of the results, we developed PeptOlink, a publicly available interactive R Shiny app (https://peptolink.serve.scilifelab.se/app/peptolink). PeptOlink allows users to visualize peptides quantified by MS on the protein sequence, along with their correlation with the corresponding Olink assay. It also provides filtering options for various peptide- and protein-level data quality factors and cross-platform correlation metrics, to aid in identifying proteins of interest. Below, we provide a few representative examples to illustrate the utility of PeptOlink in exploring cross-platform correlations: AMBP (α-1-Microglobulin/Bikunin Precursor), HYOU1 (Hypoxia Upregulated Protein 1), and MASP1 (Mannan-binding lectin Serine Protease 1). These proteins exhibited substantial variation in peptide-Olink correlations across different regions of their respective protein sequences.\u003c/p\u003e\n\u003cp\u003eThe \u003cem\u003eAMBP\u003c/em\u003e gene encodes a precursor protein that is cleaved into α1-Microglobulin (residues 20–203) and Inter-α-Trypsin (IαI) Inhibitor Light Chain (residues 206–352). Notably, peptides mapping to the α1-Microglobulin region showed stronger correlations with the AMBP Olink assay (median ρ = 0.53) compared to peptides mapping to the IαI Light Chain region (median ρ = 0.07) (Fig.\u0026nbsp;7A, Figure S20), suggesting that the Olink assay primarily measures α1-Microglobulin. For HYOU1, we observed two regions with differing cross-platform correlations. One region, shared between isoforms 1 and 2 (residues 88–602), demonstrated poor agreement between MS and Olink measurements (median ρ = 0.21), whereas the other, unique to isoform 1 (residues 647–999), exhibited moderate agreement (median ρ = 0.64) between MS and Olink measurements (Fig.\u0026nbsp;7B, Figure S20). These findings suggest that MS and Olink may have detected different isoforms of HYOU1. Finally, MASP1 had multiple isoforms in the MS data (UniProt IDs: P48740, P48740-2, P48740-3) and evidence of differing MS-Olink correlation between these isoforms. At the protein level, isoforms 1 and 3 had poor correlation with the corresponding Olink assay (ρ = 0.04 and ρ = 0.23), while isoform 2, also known as MASP3, had a moderate correlation with Olink (ρ = 0.57). At the peptide level, regions mapping uniquely to isoform 2 exhibited higher cross-platform correlations than peptides mapping to other regions (Fig.\u0026nbsp;7C, Figure S20). These findings suggest that Olink’s antibodies may be binding sequences specific to isoform 2 of MASP1.\u003c/p\u003e\n\u003cp\u003eIn summary, these examples illustrate how peptide-level analysis by MS can reveal potential differences in proteoform measurements across proteomics platforms.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we present a thorough investigation into the relative strengths and weaknesses of HiRIEF LC-MS/MS and the antibody-based Olink Explore 3072 platform, based on an in-depth analysis of the plasma proteome involving 4,362 proteins in total and 1,129 proteins overlapping between the two methods\u0026mdash;significantly more than in previous studies comparing MS and Olink\u0026rsquo;s PEAs (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). We show that these platforms exhibit highly complementary proteome coverage, high precision, concordance and complementarity in DAA, and largely good, albeit variable, quantitative agreement in protein levels. Consequently, we provide orthogonal validation for a large proportion of the targets of the Olink Explore 3072 platform. Furthermore, we identify technical factors and protein properties influencing the quantitative agreement and compare estimates of cross-platform agreement across previous studies. Finally, we demonstrate how a peptide-level analysis of cross-platform correlations can reveal insights into differences in proteoform measurement.\u003c/p\u003e \u003cp\u003eOur findings demonstrate the complementarity of in-depth MS and Olink Explore 3072 in providing a more comprehensive analysis of the plasma proteome than either approach on their own. This is in line with previous reports from studies comparing other MS workflows and the smaller Olink Target panels (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Together, these technologies offer broad proteome coverage across a wide concentration range and distinct biological processes, increasing the potential for biologically and clinically relevant discoveries. The differences in proteome coverage suggest that each platform may offer distinct advantages in specific study contexts. For example, the relatively higher coverage of dedicated plasma proteins and metabolic proteins observed for MS may be advantageous for investigating cardiovascular disease, metabolic disorders, or any condition with large effects on the plasma proteome, such as systemic infectious or inflammatory diseases. Olink\u0026rsquo;s various Explore panels, which target many low-abundance tissue-leakage and signaling proteins, may offer an advantage in the study of conditions affecting specific tissues or the immune system, for example cancer, neurological disorders, or chronic inflammatory diseases.\u003c/p\u003e \u003cp\u003eBeyond proteome coverage, these technologies have additional complementary strengths and limitations. The Olink Explore platform currently offers higher sensitivity and throughput than MS, enabling large-scale studies with high proteome coverage, though limited to pre-defined proteins. In contrast, the untargeted nature of MS provides a significant advantage for discovery-focused studies, allowing for a general characterization of the plasma proteome across different conditions and for the detection of novel proteins and PTMs. This advantage will likely become even more pronounced as sensitivity and proteome coverage continue to improve. In particular, the Orbitrap Astral has shown promising results in plasma proteome profiling, providing higher sample throughput while maintaining or even increasing profiling depth (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e). Ultimately, the choice of profiling platform should be guided by the specific goals and disease context of the study. Since both methods provide valuable insights into different, although somewhat overlapping, aspects of the plasma proteome, combining in-depth MS with Olink panels targeting relevant low-abundance proteins offers an attractive approach to plasma proteome profiling that leverages the strengths of both technologies and can be tailored to the specific application.\u003c/p\u003e \u003cp\u003eThe complementarity of in-depth MS and Olink Explore 3072 was further reflected in the analysis of protein level differences between males and females. A greater number of DAPs were detected when combining both platforms, with DAPs unique to each platform stemming from different concentration ranges of the plasma proteome. The larger log2-fold change values and the greater number of DAPs identified by Olink, particularly in the low-abundance concentration range, likely reflect the sensitivity of the PEA technology. At these low concentrations, MS might be nearing its LOD, where the biological signal could be masked by technical noise. In addition, TMT ratio compression, an artifact caused by peptide co-fragmentation in MS experiments using TMT labeling (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e), may have slightly reduced the observed differences in protein abundance between samples. Regardless of the underlying cause, larger sample sizes may be needed to reach statistical significance with MS, at least for some low-abundance proteins. Despite these limitations, MS identified numerous DAPs that were not detected by Olink, particularly in the high-abundance concentration range.\u003c/p\u003e \u003cp\u003eIn addition, the DAA exemplified how the agreement between platforms can be used to assess the validity of findings when using different profiling technologies. While only a subset of DAPs reached statistical significance with both methods, there was strong concordance in the estimated differences. Thus, because discrepancies in statistical significance could be attributed to differences in power, such as higher noise or more missing values in one platform, examining the agreement in protein level differences provides a more comprehensive evaluation of the results beyond just the significant proteins. Overall, this analysis illustrates the value of combining MS and Olink for enhancing the biological insights and increasing the number of potential biomarkers drawn from proteomic data.\u003c/p\u003e \u003cp\u003eWe identified several technical factors associated with weaker agreement between platforms, which can serve as indicators of lower measurement accuracy. These included a higher proportion of missing values, lower estimated protein concentrations, and higher technical CVs for both platforms, as well as sample QC warnings and median NPX values closer to the LOD for Olink data, and few PSMs or peptides and high precursor mass errors for MS data. In this study, we treated missing values in MS and values\u0026thinsp;\u0026lt;\u0026thinsp;LOD in Olink as equivalent, but it is important to recognize that these arise through different mechanisms. In MS workflows that employ TMT labeling, missing values are largely TMT set-specific, with peptide detection varying across runs. Detection in each run depends on a variety of biological and technical factors, not solely protein abundance or absence (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). In contrast, values\u0026thinsp;\u0026lt;\u0026thinsp;LOD in Olink assays more certainly indicate that the protein concentration was too low for reliable detection (or absent). Despite these differences, missing values had the strongest effect on lowering cross-platform correlations in both datasets, and proteins with a high proportion of missing values should therefore be handled cautiously in both types of proteomic data.\u003c/p\u003e \u003cp\u003eFurthermore, we observed that some proteins with seemingly high-quality quantification on both platforms had poor agreement between platforms. This could be explained by a multitude of factors: the measurement of different isoforms, differences introduced by sample handling or preparation, antibody cross-reactivity, PTMs, and many more. Through an analysis of cross-platform correlations at the peptide level, we were able to identify potential differential detection of proteoforms and proteolytic cleavage products. At the protein level, the examples we presented, AMBP and HYOU1, had moderate correlation between platforms. However, our results from the peptide-level analysis demonstrate that the correlations could be stronger if both platforms were measuring the same protein regions or isoforms. These examples highlight an ongoing challenge in proteomics: shifting from a gene- or protein-centric to a proteoform-centric analysis of the proteome. While MS offers some ability to distinguish between proteoforms, there is still a need to better resolve the specific epitopes and isoforms targeted by affinity binders. Previous studies have made significant efforts to validate antibody specificity and selectivity using MS (\u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e), and similar work will be needed to evaluate the affinity binders used in commercial plasma proteomics platforms such as Olink and SomaScan. Orthogonal validation with MS, as performed in the present study, represents one important pillar of affinity binder validation (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe cross-platform correlation analysis revealed comparable but slightly higher correlations between MS and Olink than have been reported in previous studies (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e), which could be explained by differences between MS workflows or differences between the Olink Target and Olink Explore panels. Our data provide evidence for the accuracy and specificity of at least one third of the Olink Explore 3072 assays with a strong correlation with HiRIEF LC-MS/MS. We also confirm with MS that the association with genetic variation\u0026mdash;the presence of pQTLs\u0026mdash;is useful to assess the accuracy of affinity-based assays. However, we also show that the pQTL approach is not without limitations, as proteins with pQTLs on both affinity-based platforms did not necessarily have a strong correlation between each other or with MS. The pQTL approach may penalize proteins that exhibit high cross-platform correlations but lack pQTLs due to an absence of genetic associations or small effect sizes. We identified a few such cases in the study by Eldjarn \u003cem\u003eet al.\u003c/em\u003e (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e), where the accuracy of the affinity-based assays was supported by our MS data despite a lack of pQTLs. Finally, through comparison with previous studies investigating cross-platform correlations between Olink and SomaScan platforms (\u003cspan additionalcitationids=\"CR16 CR17 CR18 CR19 CR20 CR21\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e), our data may also indirectly provide support for the specificity of some SomaScan assays. However, due to the lack of SomaScan data in this study, we cannot draw definitive conclusions.\u003c/p\u003e \u003cp\u003eSome limitations of this study should be considered. First, the CV calculations were based on a small number of duplicate samples, leading to uncertainty in the CV estimates. Additionally, we were unable to calculate intra- and inter-assay CVs for both technologies, and therefore, we compared the inter-assay CVs of HiRIEF LC-MS/MS to the intra-assay CVs of Olink Explore 3072. Second, some conclusions drawn from this study may not be fully generalizable to all other MS workflows and Olink platforms. Results could vary due to differences in sample preparation protocols, instruments, and data processing of MS workflows, or due to differences in affinity probes and performance characteristics of Olink platforms. These variations are particularly likely to affect the quantitative agreement between MS and Olink, as seen in the comparison of cross-platform correlations between studies. Finally, some results could differ between patient cohorts, due to the influence of factors such as age, sex, and disease state on the plasma proteome. Nevertheless, the main conclusions regarding the relative strengths and limitations of in-depth MS workflows and Olink platforms are likely to remain consistent. Future studies should include additional technical replicates and compare other in-depth MS workflows to affinity-based profiling platforms, both Olink and SomaScan, to further evaluate these technologies.\u003c/p\u003e \u003cp\u003eIn conclusion, this study provides insights into the complementary strengths and limitations of in-depth MS and Olink Explore 3072 for plasma proteome profiling. Our analysis, encompassing a large number of overlapping proteins, a thorough investigation into factors affecting platform performance, and an analysis of cross-platform correlations at both the protein and peptide level demonstrate that overall, both platforms perform well at profiling the proteome in depth, with each focusing on different portions of the plasma proteome. The complementary proteome coverage and the variable quantitative agreement highlight the added value of combining these platforms for a more comprehensive and reliable profiling of the plasma proteome, enabling a broader characterization of different diseases and more robust biomarker discovery.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003ch2\u003eEthical approval\u003c/h2\u003e\n\u003cp\u003eThis study was approved by the regional ethical review board in Stockholm, Sweden (EPN: ref no 2014/1290–32) and conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent. \u003c/p\u003e\n\u003ch2\u003eStudy design and sample collection\u003c/h2\u003e\n\u003cp\u003eThe present study cohort consists of a retrospectively selected subset of the PEX-LC cohort, which has been described previously (\u003cem\u003e38\u003c/em\u003e). Briefly, plasma samples were collected from patients referred to the Karolinska University Hospital (KUH) in Stockholm, Sweden, for investigation of suspected lung cancer between September 2014 and November 2015. The plasma samples were collected during the participants’ first visit to KUH, before diagnosis and treatment. Blood was drawn into EDTA tubes, centrifuged at 2500×g at RT for 10 min, and the resulting plasma samples were biobanked and stored at -80℃. The present study cohort includes 114 patients, with an equal number of patients diagnosed with either lung cancer or no cancer, \u003cem\u003ei.e.\u003c/em\u003e other benign lung conditions. Samples from all 114 patients were analyzed using MS, and a subset of 88 plasma samples with Olink Explore 3072. For the MS analysis, six samples were run in duplicate (aliquoted at the start of sample preparation), resulting in a total of 120 samples.\u003c/p\u003e\n\u003ch2\u003ePlasma proteome profiling\u003c/h2\u003e\n\u003ch3\u003eMass spectrometry-based proteome profiling\u003c/h3\u003e\n\u003ch4\u003ePlasma depletion and in-solution digestion\u003c/h4\u003e\n\u003cp\u003eTo reduce sample complexity and increase the number of protein identifications, the 14 most abundant plasma proteins were depleted from the samples using High Select Top14 Abundant Protein Depletion Mini Spin Columns (Themo Scientific). 10 μL of plasma was applied to each column, the columns were incubated at room temperature for 20 min with gentle end-over-end mixing, and depleted flowthroughs were obtained through centrifugation. The sample buffer was exchanged to 50 mM HEPES (pH 7.6) using 5 kDa spin concentrators (5K MWCO, 4 mL, Agilent Technologies, 5185-5987) by centrifuging three times at 5000 rpm for 30 min. Protein concentration was measured using the Micro BCA Protein Assay Kit (Thermo Scientific, 23235) to estimate the total protein amount per sample. \u003c/p\u003e\n\u003cp\u003eNext, proteins were digested into peptides using lysC and trypsin (sequencing grade modified, Pierce) following a previously described in-solution digestion protocol (\u003cem\u003e39\u003c/em\u003e). In brief, 40 μg of protein from each sample was alkylated with 8 mM chloroacetamide. 100 μL of lysC buffer (0.5 M Urea, 50 mM HEPES, pH 7.6 and 1:50 enzyme-to-protein ratio) was added and the samples were incubated overnight. The same procedure was repeated for trypsin; 100 μL of trypsin buffer (50 mM HEPES, pH 7.6, 1:50 enzyme-to-protein ratio) was added and the mixtures were incubated overnight. Finally, the samples were dried in a SpeedVac and resuspended in 50 μL TEAB pH 8.5 to a final concentration of 100 mM.\u003c/p\u003e\n\u003ch4\u003eTMT labeling\u003c/h4\u003e\n\u003cp\u003e40 μg of peptides from each sample was labelled with isobaric TMTs (TMTpro 16plex Label Reagent Set, Thermo Scientific) according to the manufacturer’s protocol. A total of 120 samples (114 distinct plasma samples, and six pairs of technical replicates) were labeled with eight sets of TMTpro 16plex, with one internal standard per set. The master pool of internal standards was made by pooling a small amount of protein from each sample, and the master pool was then split into eight internal standards of 40 μg of protein each. The TMT labeling scheme is shown in Table S15. \u003c/p\u003e\n\u003cp\u003eThe TMT labeling efficiency was determined by LC-MS/MS prior to pooling of the samples. For this, 1 μL of each sample of a TMT set was mixed, dried down, and resuspended in 10 μL of mobile phase A. Approximately 2 μg was injected into the LC-MS/MS system and analyzed with a 3h gradient. After confirming a labeling efficiency \u0026gt;95%, samples of the same TMT set were pooled. The eight resulting TMT pools were purified through solid phase extraction using SPE strata-X-C columns (Phenomenex), and purified samples were dried in a SpeedVac.\u003c/p\u003e\n\u003ch4\u003eHigh-resolution isoelectric focusing\u003c/h4\u003e\n\u003cp\u003eTo further reduce sample complexity, pooled samples were pre-fractionated using HiRIEF, following a previously described protocol (\u003cem\u003e40\u003c/em\u003e). Briefly, peptides were separated by their isoelectric point through immobilized pH gradient isoelectric focusing (IPG-IEF) on gel strips with a 3–10 pH gradient. After IEF, each gel strip was split into 72 fractions, and proteins from each fraction were eluted and transferred to a 96-well microtiter plate using a liquid-handling robot (GE Healthcare prototype). Finally, the fractionated samples were dried in a SpeedVac and stored at -20℃ until analysis with LC-MS/MS.\u003c/p\u003e\n\u003ch4\u003eLC-MS/MS\u003c/h4\u003e\n\u003cp\u003eOnline LC-MS was performed as previously described (\u003cem\u003e40\u003c/em\u003e) using a Dionex UltiMate™ 3000 RSLCnano System coupled to a Q-Exactive-HF mass spectrometer (Thermo Scientific). The contents of each plate well were dissolved in 20 uL of solvent A and 10 uL was injected. Samples were trapped on a C18 guard-desalting column (Acclaim PepMap 100, 75μm x 2 cm, nanoViper, C18, 5 µm, 100Å), and separated on a 50 cm long C18 column (Easy spray PepMap RSLC, C18, 2 μm, 100Å, 75 μm x 50 cm). The nano capillary solvent A consisted of 94.9 % water, 5 % DMSO, and 0.1 % formic acid, and solvent B consisted of 5 % water, 5 % DMSO, 89.9 % acetonitrile, and 0.1 % formic acid. At a constant flow of 0.25 μL/min, the curved gradient went from 6-10 % solvent B up to 40 % solvent B in each fraction in a dynamic range of gradient length (see Table S16), followed by a steep increase to 100% solvent B in 5 min. \u003c/p\u003e\n\u003cp\u003eFTMS (Fourier transform mass spectrometry) master scans with 60 000 resolution and mass range 300-1500 m/z were followed by data-dependent MS/MS with a resolution of 30 000 on the top 5 ions using higher energy collision dissociation (HCD) at 30% normalized collision energy. Precursors were isolated with a 2 m/z window. Automatic gain control (AGC) targets were 1\u003csup\u003e6\u003c/sup\u003e for MS1 and 1\u003csup\u003e5\u003c/sup\u003e for MS2. Maximum injection times were 100 ms for MS1 and 400 ms for MS2. The entire duty cycle lasted ~2.5 s. Dynamic exclusion was used with 30 s duration. Precursors with unassigned charge state or charge state 1 were excluded. An underfill ratio of 1% was used.\u003c/p\u003e\n\u003ch4\u003eProtein identification and quantification\u003c/h4\u003e\n\u003cp\u003eOrbitrap raw MS/MS files were converted to mzML format using msConvert from the ProteoWizard tool suite (\u003cem\u003e41\u003c/em\u003e). Spectra were searched using the ddamsproteomics Nextflow (v22.10.5) (\u003cem\u003e42\u003c/em\u003e) pipeline (https://github.com/lehtiolab/ddamsproteomics, v2.11), which runs MSGF+ (v2020.03.14) (\u003cem\u003e43\u003c/em\u003e) and Percolator (v3.04.0) (\u003cem\u003e44\u003c/em\u003e) for peptide identification. All searches were performed against a database of all human proteins from the UniProtKB/Swiss-Prot release of May 2022. MSGF+ settings included precursor mass tolerance of 10 ppm, fully tryptic peptides, a maximum peptide length of 50 amino acids and a maximum charge of 6. Fixed modifications included carbamidomethylation on cysteine residues and TMTpro 16plex on lysine residues and peptide N-termini. A variable modification was used for oxidation on methionine residues. PSMs found at 1% false discovery rate (FDR) were used to infer protein identities.\u003c/p\u003e\n\u003cp\u003eTMTpro 16plex reporter ions were quantified using OpenMS project's IsobaricAnalyzer (v2.5.0) (\u003cem\u003e45\u003c/em\u003e). Relative quantification was calculated on peptide and protein level based on PSMs mapping to only one protein group (UniProt ID) and with 1% FDR. PSMs with missing values in any channel within a TMT set were excluded. Relative quantification values were calculated for each TMT channel as the median of PSM ratios (channel/internal standard). To obtain these PSM ratios, PSM intensities were log2-transformed, and the transformed PSM intensities of the internal standard were subtracted from the transformed PSM intensities of the channel. Peptide/protein quantification values were then normalized by subtracting the median of the channel from each value. Protein FDRs were calculated using the picked-FDR method using UniProt IDs as protein groups and limited to 1% FDR.\u003c/p\u003e\n\u003ch2\u003eAntibody-based proteome profiling\u003c/h2\u003e\n\u003cp\u003eThe samples were analyzed using the Olink Proteomics PEA Explore 3072 at SciLifeLab Affinity Proteomics unit at Uppsala University and the National Genomics Infrastructure Uppsala. The detailed protocol for PEA Explore has been previously described by Wik and colleagues (\u003cem\u003e14\u003c/em\u003e).\u003c/p\u003e\n\u003cp\u003eIn summary, Olink’s PEA Explore technology utilizes pairs of antibodies conjugated with single-stranded DNA oligonucleotide reporter molecules, known as probes, which bind to their respective targets if present in the sample. When both probes in a pair bind to their target in proximity, double-stranded DNA amplicons are generated. The Explore 3072 assay comprises eight distinct 384-plex panels targeting inflammation, oncology, cardiometabolic, and neurology proteins, covering a total of 2,923 human proteins. Four of these are version I panels (\u003cem\u003ee.g.\u003c/em\u003e Inflammation), which were part of the earlier Olink Explore 1536 platform, while the remaining four are version II panels (\u003cem\u003ee.g.\u003c/em\u003e Inflammation II), added to the Olink Explore 3072 platform.\u003c/p\u003e\n\u003cp\u003eFollowing the initial probe-based immune reaction step in the PEA Explore workflow, the amplicons were extended and amplified in a two-step process, with individual sample index sequences added during the second step. After pooling the samples, the libraries were prepared and sequenced on a NovaSeq 6000 instrument (Illumina, San Diego, CA, USA). The raw BCL files were converted into count files, which were then translated into NPX values through a QC and normalization process incorporating internal and external controls, as specified by the manufacturer. In this process, QC is performed for each assay (protein) measured in each sample and for each assay overall. If the QC for an assay in a specific sample fails, the measurement receives a sample QC warning. If the overall assay QC fails, the assay receives an assay QC warning (across all samples).\u003c/p\u003e\n\u003cp\u003eThe NPX data are presented on a log2 scale, where an increase of one NPX unit corresponds to a doubling of the protein content. A high NPX value indicates a high protein concentration. Each measured protein has a LOD determined at run time based on negative controls. Values \u0026lt;LOD were retained in the analyses, unless stated otherwise, but considered missing values.\u003c/p\u003e\n\u003ch3\u003eStatistical analysis\u003c/h3\u003e\n\u003ch4\u003eProtein overlaps and reference plasma proteome\u003c/h4\u003e\n\u003cp\u003eThe reference plasma proteome was compiled from proteins listed in the Human Plasma section of the PeptideAtlas (build 2023-04, www.peptideatlas.org) (\u003cem\u003e26\u003c/em\u003e), proteins with an estimated blood concentration in the HPA (v24, www.proteinatlas.org) (\u003cem\u003e27\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e), and proteins classified as secreted to blood in the HPA, resulting in a reference set of 4,889 unique proteins based on UniProt IDs. Some proteins in the HPA data lacked UniProt IDs. Where possible, these were assigned UniProt IDs from the MS and Olink data through gene name matching. Proteins overlapping between MS, Olink, PeptideAtlas, and HPA datasets were identified by exact UniProt ID matching.\u003c/p\u003e\n\u003ch4\u003eEstimated concentration in blood\u003c/h4\u003e\n\u003cp\u003eThe estimated concentration in the blood of each protein was obtained from the HPA (\u003cem\u003e28\u003c/em\u003e) and converted to ng/mL. The concentrations in the HPA are derived from immunoassay data from the literature and MS data from the PeptideAtlas (\u003cem\u003e26\u003c/em\u003e). For proteins that had both immunoassay- and MS-based concentrations available, the MS-based concentration was used. For proteins lacking an MS-based concentration, the immunoassay-based concentration was used, if available. \u003c/p\u003e\n\u003ch4\u003eComparison of protein annotations\u003c/h4\u003e\n\u003cp\u003eThe frequency of specific HPA annotations was calculated among all proteins detected in at least one sample by MS or Olink. Overrepresentation of these annotations was tested using a hypergeometric test, with all proteins detected by MS and/or Olink used as the background (N = 4,362). P-values were adjusted for multiple testing using the FDR method, with a significance threshold of FDR \u0026lt; 0.05. The “Enriched tissue” category refers to proteins annotated as “Tissue enriched” in the Tissue section of the HPA (\u003cem\u003e27\u003c/em\u003e), meaning their mRNA expression was at least four-fold higher in a specific tissue compared to all other tissues. \u003c/p\u003e\n\u003cp\u003eORA of GO Biological Processes between platforms was performed using the compareCluster function in the clusterProfiler (version 4.14.4) (\u003cem\u003e46\u003c/em\u003e) R package. All proteins detected with MS and/or Olink were used as the background protein list (N = 4,362). Fold enrichment for GO terms was calculated as the ratio of the frequency of the input proteins (GeneRatio) to the frequency of the background proteins (BgRatio) found in the respective GO term gene list, based on UniProt IDs. For Olink assays with multiple UniProt IDs, the first one was used for analysis. P-values were adjusted for multiple testing using the FDR method, with a significance threshold of FDR \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003eThe coverage of FDA-approved plasma protein biomarkers was based on a list compiled by Anderson (\u003cem\u003e29\u003c/em\u003e). Proteins were matched between this list and the MS and Olink datasets based on UniProt IDs. Ten markers with no UniProt ID were excluded from the analysis. \u003c/p\u003e\n\u003ch4\u003eTechnical coefficients of variation\u003c/h4\u003e\n\u003cp\u003eTechnical CVs were calculated per protein and duplicate sample using the CV formula for data on a log\u003csub\u003e2\u003c/sub\u003e-scale (\u003cem\u003e47\u003c/em\u003e): \u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003cp\u003eFor proteins measured in multiple duplicate samples, the final technical CV was the mean of the individual duplicate CVs. Six samples were run in duplicate with MS, with each sample’s duplicates placed in different TMT sets. However, due to differences in detected proteins between TMT sets (missing values), each protein in the MS data had between zero and six duplicates available for the CV calculation. In the Olink data, technical CVs were calculated on values above the LOD of one duplicate of the Olink Sample Control, which is a standard pooled plasma sample. Because values \u0026lt;LOD were excluded, and the Sample Control failed for some assays, CVs could not be calculated for all proteins. Proteins with outlier CVs (CV \u0026gt; 100%) were excluded from the analysis. This applied to four proteins in the MS data. \u003c/p\u003e\n\u003ch4\u003eDifferential abundance analysis\u003c/h4\u003e\n\u003cp\u003eDAA was performed for overlapping proteins with no missing values (N = 569) between females (N = 37) and males (N = 51) using a two-sided Welch’s t-test, with males as the reference group. P-values were adjusted for multiple testing per platform using the FDR method, with a significance threshold of FDR \u0026lt; 0.05. The categorical agreement of the log2-fold change values was calculated as the percentage of proteins showing the same direction of change (positive or negative) in both platforms.\u003c/p\u003e\n\u003ch4\u003eCross-platform correlation analyses\u003c/h4\u003e\n\u003cp\u003eCorrelations between MS and Olink measurements of matched proteins were calculated using Spearman’s rank correlation coefficient, and all correlations were presented without filtering for statistical significance. Proteins with less than eight overlapping data points were excluded from the analyses. Correlations were calculated on a complete dataset of all overlapping proteins (N = 1129), a cleaned dataset where values \u0026lt;LOD and QC warnings in the Olink data were set to missing (N = 1064), and cleaned datasets additionally filtered for a maximum of 50% missing values (N = 791) or no missing values (N = 463). The cross-platform correlations were divided into categories of no correlation: r ∈ [-1, 0.3); weak correlation: r ∈ [0.3, 0.5); moderate correlation: r ∈ [0.5, 0.7); and strong correlation: r ∈ [0.7, 1.0].\u003c/p\u003e\n\u003cp\u003eAssociations between the MS-Olink correlations and technical factors were assessed using both Spearman’s and Pearson’s correlation coefficients, at a significance level of α = 0.05. Information on protein mass, length, and number of isoforms were obtained from UniProt release 2023_03, and protein concentration from the HPA as described above. All other technical factors analyzed were obtained or calculated from the MS and Olink data files.\u003c/p\u003e\n\u003cp\u003eFor the peptide-level analysis, MS peptides were matched to Olink assays based on gene name (\u003cem\u003ei.e. \u003c/em\u003eOlink assay name), and correlations were calculated using Spearman’s rank correlation coefficient. Peptides quantified in less than 15 samples, and genes with less than two peptides were excluded from the analysis. Information on the sequence positions of different isoforms and cleavage products of AMBP, HYOU1, and MASP1 were obtained manually from UniProt (release 2025_01, https://www.uniprot.org/). The fasta sequences of proteins included in the PeptOlink R Shiny app were downloaded from UniProt (release 2022_05). \u003c/p\u003e\n\u003ch4\u003eEnrichment analyses\u003c/h4\u003e\n\u003cp\u003eGSEAs of GO terms, HPA annotations, and MSigDB, KEGG, and Reactome gene sets were performed using the clusterProfiler R package, with a ranked list of the MS-Olink correlations of all overlapping proteins as the input (N = 1129). P-values were adjusted for multiple testing using the FDR method, with a significance threshold of FDR \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003eORAs of GO terms, HPA annotations, and MSigDB, KEGG, and Reactome gene sets among proteins in the low correlation (ρ \u0026lt; 0.3) and strong correlation (ρ ≥ 0.7) categories were performed as described for the comparison of GO Biological Processes between platforms. All overlapping proteins (N = 1129) were used as the background.\u003c/p\u003e\n\u003ch4\u003eComparison with previous studies\u003c/h4\u003e\n\u003cp\u003eThe previous studies included in the comparison of cross-platform correlations are summarized in Table S17. Cross-platform correlations were obtained from the supplementary materials of all publications except Petrera \u003cem\u003eet al\u003c/em\u003e. (\u003cem\u003e24\u003c/em\u003e), for which MS and Olink data files were downloaded, and correlations were calculated between the DDA-MS and Olink measurements of all overlapping proteins matched by UniProt IDs. Proteins were matched between studies primarily by UniProt IDs, or by gene name if UniProt IDs were not provided. Several of the studies comparing Olink and SomaScan calculated Olink-SomaScan correlations both on normalized and non-normalized SomaScan data. For these studies, the correlations calculated on the normalized data were used in the cross-study comparison. \u003c/p\u003e\n\u003cp\u003eFor the comparison of correlations by confidence tiers defined in Eldjarn \u003cem\u003eet al. \u003c/em\u003e(\u003cem\u003e15\u003c/em\u003e), data were obtained from Supplementary Table 29 of the original publication. The authors defined the confidence tiers as follows: tier 1, the highest confidence tier, included proteins with an Olink-SomaScan correlation \u0026gt;0.5 and \u003cem\u003ecis\u003c/em\u003e-pQTLs detected on both platforms; tier 2 included proteins with a correlation of ≤0.5 and a \u003cem\u003ecis\u003c/em\u003e-pQTL detected on both platforms; and tier 3 included proteins with a \u003cem\u003ecis\u003c/em\u003e-pQTLs identified on only one or neither platform.\u003c/p\u003e\n\u003ch4\u003eSoftware\u003c/h4\u003e\n\u003cp\u003eAll statistical analyses were performed in R (version 4.4.2) (\u003cem\u003e48\u003c/em\u003e). Figures were assembled in Adobe Illustrator 2025 (version 29.3.1).\u003c/p\u003e\n\u003ch4\u003eData visualization\u003c/h4\u003e\n\u003cp\u003eData visualization was performed in R using the ggplot2, ggpubr, ggpp, ggExtra, ggrepel, patchwork, RColorBrewer, and colorspace packages. Venn diagrams were generated using the eulerr R package. Point densities were calculated and visualized using the ggpointdensity function in the ggpointdensity package. The geom_smooth function from the ggplot2 R package was used to add regression lines to scatter plots. The peptide correlation plots were created using the plotly and heatmaply R packages. \u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eData availability\u003c/p\u003e\n\u003cp\u003eThe mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD061144. The affinity proteomics data have been deposited to the PRIDE repository with the dataset identifier PAD000006. All other data supporting the findings of this study are available within the paper and its supplementary information files.\u003c/p\u003e\n\u003cp\u003eCode availability\u003c/p\u003e\n\u003cp\u003eThe R code used for the analyses is deposited at https://github.com/noorasissala/MS-Olink-comparison. The code for the PeptOlink R Shiny app is deposited at https://github.com/isabelle-leo/PeptOlink.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eWe acknowledge support from the Global Proteomics and Proteogenomics Unit and the Affinity Proteomics Unit at the Science for Life Laboratory, as well as the National Genomics Infrastructure Uppsala. The project was funded by the Science for Life Laboratory Technology Development Grant 2022 (MP).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAuthor contributions\u003c/p\u003e\n\u003cp\u003eConceptualization: MP, M\u0026Aring;, CF, NS. Data Curation: JF. Formal Analysis: NS, IL. Funding Acquisition: MP, M\u0026Aring;, CF, JL, LEE. Methodology: MP, M\u0026Aring;, CF, NS, IL, HB. Resources: MP, M\u0026Aring;, CF, LEE, JL. Software: IL, NS. Investigation: XC, M\u0026Aring;. Visualization: NS, IL. Supervision: MP, JL, HB, LEE. Writing\u0026mdash;original draft: NS, MP. Writing\u0026mdash;review \u0026amp; editing: NS, HB, IL, XC, JF, LEE, JL, CF, M\u0026Aring;, MP.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eB. He, Z. Huang, C. Huang, E. C. Nice, Clinical applications of plasma proteomics and peptidomics: Towards precision medicine. \u003cem\u003ePROTEOMICS \u0026ndash; Clin. Appl.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 2100097 (2022).\u003c/li\u003e\n\u003cli\u003eY. Liu, A. Beyer, R. Aebersold, On the Dependency of Cellular Protein Levels on mRNA Abundance. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e165\u003c/strong\u003e, 535\u0026ndash;550 (2016).\u003c/li\u003e\n\u003cli\u003eP. E. Geyer, L. M. Holdt, D. Teupser, M. Mann, Revisiting biomarker discovery by plasma proteomics. \u003cem\u003eMol. Syst. Biol.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 942 (2017).\u003c/li\u003e\n\u003cli\u003eN. L. Anderson, N. G. Anderson, The Human Plasma Proteome: History, Character, and Diagnostic Prospects*. \u003cem\u003eMol. Cell. Proteomics\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 845\u0026ndash;867 (2002).\u003c/li\u003e\n\u003cli\u003eL. Dayon, O. Cominetti, M. Affolter, Proteomics of human biological fluids for biomarker discoveries: technical advances and recent applications. \u003cem\u003eExpert Rev. Proteomics\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 131\u0026ndash;151 (2022).\u003c/li\u003e\n\u003cli\u003eJ. M. Bader, V. Albrecht, M. Mann, MS-Based Proteomics of Body Fluids: The End of the Beginning. \u003cem\u003eMol. Cell. Proteomics\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 100577 (2023).\u003c/li\u003e\n\u003cli\u003eM. Pernemalm, A. Sandberg, Y. Zhu, J. Boekel, D. Tamburro, J. M. Schwenk, A. Bj\u0026ouml;rk, M. Wahren-Herlenius, H. \u0026Aring;mark, C.-G. \u0026Ouml;stenson, M. Westgren, J. Lehti\u0026ouml;, In-depth human plasma proteome analysis captures tissue proteins and transfer of protein variants across the placenta. \u003cem\u003eeLife\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, e41608 (2019).\u003c/li\u003e\n\u003cli\u003eH. Babačić, W. Christ, J. E. Ara\u0026uacute;jo, G. Mermelekas, N. Sharma, J. Tynell, M. Garc\u0026iacute;a, R. Varnaite, H. Asgeirsson, H. Glans, J. Lehti\u0026ouml;, S. Gredmark-Russ, J. Klingstr\u0026ouml;m, M. Pernemalm, Comprehensive proteomics and meta-analysis of COVID-19 host response. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 5921 (2023).\u003c/li\u003e\n\u003cli\u003eM. Tognetti, K. Sklodowski, S. M\u0026uuml;ller, D. Kamber, J. Muntel, R. Bruderer, L. Reiter, Biomarker Candidates for Tumors Identified from Deep-Profiled Plasma Stem Predominantly from the Low Abundant Area. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1718\u0026ndash;1735 (2022).\u003c/li\u003e\n\u003cli\u003eK. Suhre, G. R. Venkataraman, H. Guturu, A. Halama, N. Stephan, G. Thareja, H. Sarwath, K. Motamedchaboki, M. K. R. Donovan, A. Siddiqui, S. Batzoglou, F. Schmidt, Nanoparticle enrichment mass-spectrometry proteomics identifies protein-altering variants for precise pQTL mapping. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 989 (2024).\u003c/li\u003e\n\u003cli\u003eK. Suhre, M. I. McCarthy, J. M. Schwenk, Genetics meets proteomics: perspectives for large population-based studies. \u003cem\u003eNat. Rev. Genet.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 19\u0026ndash;37 (2021).\u003c/li\u003e\n\u003cli\u003eB. B. Sun, J. Chiou, M. Traylor, C. Benner, Y.-H. Hsu, T. G. Richardson, P. Surendran, A. Mahajan, C. Robins, S. G. Vasquez-Grinnell, L. Hou, E. M. Kvikstad, O. S. Burren, J. Davitte, K. L. Ferber, C. E. Gillies, \u0026Aring;. K. Hedman, S. Hu, T. Lin, R. Mikkilineni, R. K. Pendergrass, C. Pickering, B. Prins, D. Baird, C.-Y. Chen, L. D. Ward, A. M. Deaton, S. Welsh, C. M. Willis, N. Lehner, M. Arnold, M. A. W\u0026ouml;rheide, K. Suhre, G. Kastenm\u0026uuml;ller, A. Sethi, M. Cule, A. Raj, H. M. Kang, L. Burkitt-Gray, E. Melamud, M. H. Black, E. B. Fauman, J. M. M. Howson, H. M. Kang, M. I. McCarthy, P. Nioi, S. Petrovski, R. A. Scott, E. N. Smith, S. Szalma, D. M. Waterworth, L. J. Mitnaul, J. D. Szustakowski, B. W. Gibson, M. R. Miller, C. D. Whelan, Plasma proteomic associations with genetics and health in the UK Biobank. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e622\u003c/strong\u003e, 329\u0026ndash;338 (2023).\u003c/li\u003e\n\u003cli\u003eA. M. Shah, P. L. Myhre, V. Arthur, P. Dorbala, H. Rasheed, L. F. Buckley, B. Claggett, G. Liu, J. Ma, N. Q. Nguyen, K. Matsushita, C. Ndumele, A. Tin, K. Hveem, C. Jonasson, H. Dalen, E. Boerwinkle, R. C. Hoogeveen, C. Ballantyne, J. Coresh, T. Omland, B. Yu, Large scale plasma proteomics identifies novel proteins and protein networks associated with heart failure development. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 528 (2024).\u003c/li\u003e\n\u003cli\u003eL. Wik, N. Nordberg, J. Broberg, J. Bj\u0026ouml;rkesten, E. Assarsson, S. Henriksson, I. Grundberg, E. Pettersson, C. Westerberg, E. Liljeroth, A. Falck, M. Lundberg, Proximity Extension Assay in Combination with Next-Generation Sequencing for High-throughput Proteome-wide Analysis. \u003cem\u003eMol. Cell. Proteomics\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 100168 (2021).\u003c/li\u003e\n\u003cli\u003eG. H. Eldjarn, E. Ferkingstad, S. H. Lund, H. Helgason, O. T. Magnusson, K. Gunnarsdottir, T. A. Olafsdottir, B. V. Halldorsson, P. I. Olason, F. Zink, S. A. Gudjonsson, G. Sveinbjornsson, M. I. Magnusson, A. Helgason, A. Oddsson, G. H. Halldorsson, M. K. Magnusson, S. Saevarsdottir, T. Eiriksdottir, G. Masson, H. Stefansson, I. Jonsdottir, H. Holm, T. Rafnar, P. Melsted, J. Saemundsdottir, G. L. Norddahl, G. Thorleifsson, M. O. Ulfarsson, D. F. Gudbjartsson, U. Thorsteinsdottir, P. Sulem, K. Stefansson, Large-scale plasma proteomics comparisons through genetics and disease associations. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e622\u003c/strong\u003e, 348\u0026ndash;358 (2023).\u003c/li\u003e\n\u003cli\u003eD. H. Katz, J. M. Robbins, S. Deng, U. A. Tahir, A. G. Bick, A. Pampana, Z. Yu, D. Ngo, M. D. Benson, Z.-Z. Chen, D. E. Cruz, D. Shen, Y. Gao, C. Bouchard, M. A. Sarzynski, A. Correa, P. Natarajan, J. G. Wilson, R. E. Gerszten, Proteomic profiling platforms head to head: Leveraging genetics and clinical traits to compare aptamer- and antibody-based methods. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, eabm5164 (2022).\u003c/li\u003e\n\u003cli\u003eM. R. Rooney, J. Chen, C. M. Ballantyne, R. C. Hoogeveen, O. Tang, M. E. Grams, A. Tin, C. E. Ndumele, F. Zannad, D. J. Couper, W. Tang, E. Selvin, J. Coresh, Comparison of Proteomic Measurements Across Platforms in the Atherosclerosis Risk in Communities (ARIC) Study. \u003cem\u003eClin. Chem.\u003c/em\u003e \u003cstrong\u003e69\u003c/strong\u003e, 68\u0026ndash;79 (2023).\u003c/li\u003e\n\u003cli\u003eL. M. Raffield, H. Dang, K. A. Pratte, S. Jacobson, L. A. Gillenwater, E. Ampleford, I. Barjaktarevic, P. Basta, C. B. Clish, A. P. Comellas, E. Cornell, J. L. Curtis, C. Doerschuk, P. Durda, C. Emson, C. M. Freeman, X. Guo, A. T. Hastie, G. A. Hawkins, J. Herrera, W. C. Johnson, W. W. Labaki, Y. Liu, B. Masters, M. Miller, V. E. Ortega, G. Papanicolaou, S. Peters, K. D. Taylor, S. S. Rich, J. I. Rotter, P. Auer, A. P. Reiner, R. P. Tracy, D. Ngo, R. E. Gerszten, W. K. O\u0026rsquo;Neal, R. P. Bowler, NHLBI Trans-Omics for Precision Medicine (TOPMed) Consortium, Comparison of Proteomic Assessment Methods in Multiple Cohort Studies. \u003cem\u003eProteomics\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, e1900278 (2020).\u003c/li\u003e\n\u003cli\u003eM. Pietzner, E. Wheeler, J. Carrasco-Zanini, A. Cortes, M. Koprulu, M. A. W\u0026ouml;rheide, E. Oerton, J. Cook, I. D. Stewart, N. D. Kerrison, J. Luan, J. Raffler, M. Arnold, W. Arlt, S. O\u0026rsquo;Rahilly, G. Kastenm\u0026uuml;ller, E. R. Gamazon, A. D. Hingorani, R. A. Scott, N. J. Wareham, C. Langenberg, Mapping the proteo-genomic convergence of human diseases. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e374\u003c/strong\u003e, eabj1541 (2021).\u003c/li\u003e\n\u003cli\u003eF. Finkernagel, S. Reinartz, M. Schuldner, A. Malz, J. M. Jansen, U. Wagner, T. Worzfeld, J. Graumann, E. P. von Strandmann, R. M\u0026uuml;ller, Dual-platform affinity proteomics identifies links between the recurrence of ovarian carcinoma and proteins released into the tumor microenvironment. \u003cem\u003eTheranostics\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 6601\u0026ndash;6617 (2019).\u003c/li\u003e\n\u003cli\u003eD. E. Haslam, J. Li, S. T. Dillon, X. Gu, Y. Cao, O. A. Zeleznik, N. Sasamoto, X. Zhang, A. H. Eliassen, L. Liang, M. J. Stampfer, S. Mora, Z.-Z. Chen, K. L. Terry, R. E. Gerszten, F. B. Hu, A. T. Chan, T. A. Libermann, S. N. Bhupathiraju, Stability and reproducibility of proteomic profiles in epidemiological studies: comparing the Olink and SOMAscan platforms. \u003cem\u003ePROTEOMICS\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 2100170 (2022).\u003c/li\u003e\n\u003cli\u003eE. B. Dammer, L. Ping, D. M. Duong, E. S. Modeste, N. T. Seyfried, J. J. Lah, A. I. Levey, E. C. B. Johnson, Multi-platform proteomic analysis of Alzheimer\u0026rsquo;s disease cerebrospinal fluid and plasma reveals network biomarkers associated with proteostasis and the matrisome. \u003cem\u003eAlzheimers Res. Ther.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 174 (2022).\u003c/li\u003e\n\u003cli\u003eJ. Graumann, F. Finkernagel, S. Reinartz, T. Stief, D. Br\u0026ouml;dje, H. Renz, J. M. Jansen, U. Wagner, T. Worzfeld, E. Pogge von Strandmann, R. M\u0026uuml;ller, Multi-platform Affinity Proteomics Identify Proteins Linked to Metastasis and Immune Suppression in Ovarian Cancer Plasma. \u003cem\u003eFront. Oncol.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1150 (2019).\u003c/li\u003e\n\u003cli\u003eA. Petrera, C. von Toerne, J. Behler, C. Huth, B. Thorand, A. Hilgendorff, S. M. Hauck, Multiplatform Approach for Plasma Proteomics: Complementarity of Olink Proximity Extension Assay Technology to Mass Spectrometry-Based Protein Profiling. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 751\u0026ndash;762 (2021).\u003c/li\u003e\n\u003cli\u003eA. K. Nilsson, H. Rydbeck, A. Thorsell, S. Fr\u0026auml;ndberg, H. Barreto Henriksson, C. Hesse, G. Hellgren, P. Lundgren, A. Hellstr\u0026ouml;m, The proteome signature of cord blood plasma with high hematopoietic stem and progenitor cell count. \u003cem\u003eStem Cell Res.\u003c/em\u003e \u003cstrong\u003e61\u003c/strong\u003e, 102752 (2022).\u003c/li\u003e\n\u003cli\u003eE. W. Deutsch, G. S. Omenn, Z. Sun, M. Maes, M. Pernemalm, K. K. Palaniappan, N. Letunica, Y. Vandenbrouck, V. Brun, S. Tao, X. Yu, P. E. Geyer, V. Ignjatovic, R. L. Moritz, J. M. Schwenk, Advances and Utility of the Human Plasma Proteome. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 5241\u0026ndash;5263 (2021).\u003c/li\u003e\n\u003cli\u003eM. Uhl\u0026eacute;n, L. Fagerberg, B. M. Hallstr\u0026ouml;m, C. Lindskog, P. Oksvold, A. Mardinoglu, \u0026Aring;. Sivertsson, C. Kampf, E. Sj\u0026ouml;stedt, A. Asplund, I. Olsson, K. Edlund, E. Lundberg, S. Navani, C. A.-K. Szigyarto, J. Odeberg, D. Djureinovic, J. O. Takanen, S. Hober, T. Alm, P.-H. Edqvist, H. Berling, H. Tegel, J. Mulder, J. Rockberg, P. Nilsson, J. M. Schwenk, M. Hamsten, K. von Feilitzen, M. Forsberg, L. Persson, F. Johansson, M. Zwahlen, G. von Heijne, J. Nielsen, F. Pont\u0026eacute;n, Tissue-based map of the human proteome. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e347\u003c/strong\u003e, 1260419 (2015).\u003c/li\u003e\n\u003cli\u003eM. Uhl\u0026eacute;n, M. J. Karlsson, A. Hober, A.-S. Svensson, J. Scheffel, D. Kotol, W. Zhong, A. Tebani, L. Strandberg, F. Edfors, E. Sj\u0026ouml;stedt, J. Mulder, A. Mardinoglu, A. Berling, S. Ekblad, M. Dannemeyer, S. Kanje, J. Rockberg, M. Lundqvist, M. Malm, A.-L. Volk, P. Nilsson, A. M\u0026aring;nberg, T. Dodig-Crnkovic, E. Pin, M. Zwahlen, P. Oksvold, K. von Feilitzen, R. S. H\u0026auml;ussler, M.-G. Hong, C. Lindskog, F. Ponten, B. Katona, J. Vuu, E. Lindstr\u0026ouml;m, J. Nielsen, J. Robinson, B. Ayoglu, D. Mahdessian, D. Sullivan, P. Thul, F. Danielsson, C. Stadler, E. Lundberg, G. Bergstr\u0026ouml;m, A. Gummesson, B. G. Voldborg, H. Tegel, S. Hober, B. Forsstr\u0026ouml;m, J. M. Schwenk, L. Fagerberg, \u0026Aring;. Sivertsson, The human secretome. \u003cem\u003eSci. Signal.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, eaaz0274 (2019).\u003c/li\u003e\n\u003cli\u003eN. L. Anderson, The Clinical Plasma Proteome: A Survey of Clinical Assays for Proteins in Plasma and Serum. \u003cem\u003eClin. Chem.\u003c/em\u003e \u003cstrong\u003e56\u003c/strong\u003e, 177\u0026ndash;185 (2010).\u003c/li\u003e\n\u003cli\u003eL. R. Heil, E. Damoc, T. N. Arrey, A. Pashkova, E. Denisov, J. Petzoldt, A. C. Peterson, C. Hsu, B. C. Searle, N. Shulman, M. Riffle, B. Connolly, B. X. MacLean, P. M. Remes, M. W. Senko, H. I. Stewart, C. Hock, A. A. Makarov, D. Hermanson, V. Zabrouskov, C. C. Wu, M. J. MacCoss, Evaluating the Performance of the Astral Mass Analyzer for Quantitative Proteomics Using Data-Independent Acquisition. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 3290\u0026ndash;3300 (2023).\u003c/li\u003e\n\u003cli\u003eN. G. Hendricks, S. D. Bhosale, A. J. Keoseyan, J. Ortiz, A. Stotland, S. Seyedmohammad, C. D. L. Nguyen, J. T. Bui, A. Moradian, S. M. Mockus, J. E. Van Eyk, An Inflection Point in High-Throughput Proteomics with Orbitrap Astral: Analysis of Biofluids, Cells, and Tissues. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 4163\u0026ndash;4169 (2024).\u003c/li\u003e\n\u003cli\u003eM. M. Savitski, T. Mathieson, N. Zinn, G. Sweetman, C. Doce, I. Becher, F. Pachl, B. Kuster, M. Bantscheff, Measuring and Managing Ratio Compression for Accurate iTRAQ/TMT Quantification. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 3586\u0026ndash;3598 (2013).\u003c/li\u003e\n\u003cli\u003eM. L. Gardner, M. A. Freitas, Multiple Imputation Approaches Applied to the Missing Value Problem in Bottom-Up Proteomics. \u003cem\u003eInt. J. Mol. Sci.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 9650 (2021).\u003c/li\u003e\n\u003cli\u003eC. Fredolini, S. Bystr\u0026ouml;m, L. Sanchez-Rivera, M. Ioannou, D. Tamburro, F. Pont\u0026eacute;n, R. M. Branca, P. Nilsson, J. Lehti\u0026ouml;, J. M. Schwenk, Systematic assessment of antibody selectivity in plasma based on a resource of enrichment profiles. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 8324 (2019).\u003c/li\u003e\n\u003cli\u003eA. Andersson, J. Remnest\u0026aring;l, B. Nellg\u0026aring;rd, H. Vunk, D. Kotol, F. Edfors, M. Uhl\u0026eacute;n, J. M. Schwenk, L. L. Ilag, H. Zetterberg, K. Blennow, A. M\u0026aring;nberg, P. Nilsson, C. Fredolini, Development of parallel reaction monitoring assays for cerebrospinal fluid proteins associated with Alzheimer\u0026rsquo;s disease. \u003cem\u003eClin. Chim. Acta\u003c/em\u003e \u003cstrong\u003e494\u003c/strong\u003e, 79\u0026ndash;93 (2019).\u003c/li\u003e\n\u003cli\u003eK. Sikorski, A. Mehta, M. Inngjerdingen, F. Thakor, S. Kling, T. Kalina, T. A. Nyman, M. E. Stensland, W. Zhou, G. A. de Souza, L. Holden, J. Stuchly, M. Templin, F. Lund-Johansen, A high-throughput pipeline for validation of antibodies. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 909\u0026ndash;912 (2018).\u003c/li\u003e\n\u003cli\u003eM. Uhlen, A. Bandrowski, S. Carr, A. Edwards, J. Ellenberg, E. Lundberg, D. L. Rimm, H. Rodriguez, T. Hiltke, M. Snyder, T. Yamamoto, A proposal for validation of antibodies. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 823\u0026ndash;827 (2016).\u003c/li\u003e\n\u003cli\u003eA. Levitsky, M. Pernemalm, B.-M. Bernhardson, J. Forshed, K. K\u0026ouml;lbeck, M. Olin, R. Henriksson, J. Lehti\u0026ouml;, C. Tishelman, L. E. Eriksson, Early symptoms and sensations as predictors of lung cancer: a machine learning multivariate model. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 16504 (2019).\u003c/li\u003e\n\u003cli\u003eX. Cao, A. Sandberg, J. E. Ara\u0026uacute;jo, F. Cvetkovski, E. Berglund, L. E. Eriksson, M. Pernemalm, Evaluation of Spin Columns for Human Plasma Depletion to Facilitate MS-Based Proteomics Analysis of Plasma. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 4610\u0026ndash;4620 (2021).\u003c/li\u003e\n\u003cli\u003eR. M. M. Branca, L. M. Orre, H. J. Johansson, V. Granholm, M. Huss, \u0026Aring;. P\u0026eacute;rez-Bercoff, J. Forshed, L. K\u0026auml;ll, J. Lehti\u0026ouml;, HiRIEF LC-MS enables deep proteome coverage and unbiased proteogenomics. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 59\u0026ndash;62 (2014).\u003c/li\u003e\n\u003cli\u003eM. C. Chambers, B. Maclean, R. Burke, D. Amodei, D. L. Ruderman, S. Neumann, L. Gatto, B. Fischer, B. Pratt, J. Egertson, K. Hoff, D. Kessner, N. Tasman, N. Shulman, B. Frewen, T. A. Baker, M.-Y. Brusniak, C. Paulse, D. Creasy, L. Flashner, K. Kani, C. Moulding, S. L. Seymour, L. M. Nuwaysir, B. Lefebvre, F. Kuhlmann, J. Roark, P. Rainer, S. Detlev, T. Hemenway, A. Huhmer, J. Langridge, B. Connolly, T. Chadick, K. Holly, J. Eckels, E. W. Deutsch, R. L. Moritz, J. E. Katz, D. B. Agus, M. MacCoss, D. L. Tabb, P. Mallick, A Cross-platform Toolkit for Mass Spectrometry and Proteomics. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 918\u0026ndash;920 (2012).\u003c/li\u003e\n\u003cli\u003eP. Di Tommaso, M. Chatzou, E. W. Floden, P. P. Barja, E. Palumbo, C. Notredame, Nextflow enables reproducible computational workflows. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 316\u0026ndash;319 (2017).\u003c/li\u003e\n\u003cli\u003eS. Kim, P. A. Pevzner, MS-GF+ makes progress towards a universal database search tool for proteomics. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 5277 (2014).\u003c/li\u003e\n\u003cli\u003eV. Granholm, S. Kim, J. C. F. Navarro, E. Sj\u0026ouml;lund, R. D. Smith, L. K\u0026auml;ll, Fast and Accurate Database Searches with MS-GF+Percolator. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 890\u0026ndash;897 (2014).\u003c/li\u003e\n\u003cli\u003eH. L. R\u0026ouml;st, T. Sachsenberg, S. Aiche, C. Bielow, H. Weisser, F. Aicheler, S. Andreotti, H.-C. Ehrlich, P. Gutenbrunner, E. Kenar, X. Liang, S. Nahnsen, L. Nilse, J. Pfeuffer, G. Rosenberger, M. Rurik, U. Schmitt, J. Veit, M. Walzer, D. Wojnar, W. E. Wolski, O. Schilling, J. S. Choudhary, L. Malmstr\u0026ouml;m, R. Aebersold, K. Reinert, O. Kohlbacher, OpenMS: a flexible open-source software platform for mass spectrometry data analysis. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 741\u0026ndash;748 (2016).\u003c/li\u003e\n\u003cli\u003eS. Xu, E. Hu, Y. Cai, Z. Xie, X. Luo, L. Zhan, W. Tang, Q. Wang, B. Liu, R. Wang, W. Xie, T. Wu, L. Xie, G. Yu, Using clusterProfiler to characterize multiomics data. \u003cem\u003eNat. Protoc.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 3292\u0026ndash;3320 (2024).\u003c/li\u003e\n\u003cli\u003eJ. A. Canchola, S. Tang, P. Hemyari, E. Paxinos, E. Marins, Correct use of percent coefficient of variation (%CV) formula for log-transformed data. \u003cem\u003eMOJ Proteomics Bioinforma.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 316\u0026ndash;317 (2017).\u003c/li\u003e\n\u003cli\u003eR Core Team, \u003cem\u003eR: A Language and Environment for Statistical Computing\u003c/em\u003e (R Foundation for Statistical Computing, Vienna, Austria, 2024; https://www.R-project.org/).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"Karolinska Institute","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Plasma proteomics, Mass spectrometry, Affinity proteomics, Proximity extension assay, Olink Explore 3072, Biomarker discovery","lastPublishedDoi":"10.21203/rs.3.rs-6501601/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6501601/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRecent advances in proteomics technologies have expanded the depth and scale of plasma proteome analyses, offering new opportunities for biomarker discovery and precision medicine. However, understanding the strengths and limitations of these technologies is crucial for study design and platform selection. We evaluated the performance and quantitative agreement of an in-depth mass spectrometry method (HiRIEF LC-MS/MS) and the Olink Explore 3072 proximity extension assay on 88 plasma samples, with 1,129 proteins analyzed with both methods. The technologies demonstrated complementary proteome coverage, high precision, and concordance in differential abundance analysis. Quantitative agreement in protein levels was moderate (median correlation 0.59, inter-quartile range 0.33\u0026ndash;0.75), influenced by various technical factors. In addition, we introduce PeptOlink, a public resource for analyzing peptide-level quantitative agreement. Our findings highlight the complementary strengths of mass spectrometry proteomics and Olink Explore 3072, underscoring the value of combining both for comprehensive and reliable profiling of the plasma proteome.\u003c/p\u003e","manuscriptTitle":"Comparative evaluation of in-depth mass spectrometry and Olink Explore 3072 for plasma proteome profiling","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-05 19:02:38","doi":"10.21203/rs.3.rs-6501601/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"abb4f8cf-17e9-4697-a284-94278840b5de","owner":[],"postedDate":"May 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":47858958,"name":"Mass Spectrometry"},{"id":47858959,"name":"Analytical Biochemistry"},{"id":47858960,"name":"Applied Biochemistry"},{"id":47858961,"name":"Biochemical Research Methods"},{"id":47858962,"name":"Translational Medicine"},{"id":47858963,"name":"Personalized Medicine"}],"tags":[],"updatedAt":"2026-01-26T19:42:10+00:00","versionOfRecord":{"articleIdentity":"rs-6501601","link":"https://doi.org/10.1038/s42004-025-01753-2","journal":{"identity":"communications-chemistry","isVorOnly":false,"title":"Communications Chemistry"},"publishedOn":"2025-11-04 00:00:00","publishedOnDateReadable":"November 4th, 2025"},"versionCreatedAt":"2025-05-05 19:02:38","video":"","vorDoi":"10.1038/s42004-025-01753-2","vorDoiUrl":"https://doi.org/10.1038/s42004-025-01753-2","workflowStages":[]},"version":"v1","identity":"rs-6501601","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6501601","identity":"rs-6501601","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}