The evolution of analytical techniques for multiplex analysis of protein biomarkers

Abstract

Introduction:

The landscape of biomarker development has evolved with advanced analytical technologies, particularly affinity- and mass spectrometry-based techniques. These advancements have deepened our understanding of disease mechanisms, enabling the development of precise diagnostic tools and personalized medicine. Protein biomarkers, which play pivotal roles in biological processes, have become invaluable in diagnosing and monitoring diseases, aided by their presence in various biological samples and the availability of established detection methods.

Areas covered:

This review covers the role of protein biomarkers in clinical practice, the development and dimensionality of protein biomarkers, advancements in detection technologies, a comparison of these technologies, and discusses future directions in biomarker discovery and disease mechanism elucidation.

Expert opinion:

Advances in biomarker technologies have the potential to transform diagnostics and personalized treatment but face challenges such as high costs and technical complexity. Enhancing reproducibility and integrating multi-omics approaches may offer better insights. The field should evolve towards high-throughput, automated methods, continuously adapting research and clinical practices.

1. Introduction

In biomedical research, the discovery and validation of biomarkers have emerged as pivotal processes in advancing precision medicine and improving patient outcome. Biomarkers are measurable indicators such as genes, proteins, metabolites, or other biomolecules that reflect variability influenced by genetic, environmental, and other factors. Among these, proteins are particularly valuable as biomarkers due to their central roles as primary effectors in biological processes, their ability to reveal physiological and pathological states, their accessibility across various biological samples, and the availability of established detection methods.

In clinical settings, biomarkers can be used as risk indicators, for predicting disease progression or recurrence, assessing disease status, monitoring response to treatment, and identifying toxicity of exposure[1]. For example, the Cardio IQ panel assesses cardiovascular risk by analyzing factors such as cholesterol, inflammation, and lipoproteins, while Troponin and B-type Natriuretic Peptide (BNP) are crucial for diagnosing acute myocardial infarction and heart failure, respectively[2]. In instances of infectious diseases, the Sepsis Panel includes biomarkers such as Procalcitonin and C-reactive protein (CRP) to diagnose sepsis[3,4]. The OVA1 test for ovarian cancer includes a 5-protein panel[5]. These panels, and many more, serve as crucial tools in clinical practice, enabling precision medicine by not only diagnosing diseases but also developing patient-specific treatment strategies. Biomarker panels help select effective treatment options, reduce overtreatment, and maintain the quality of life for patients.

With these promises, proteins have received increased interest in biomarker studies. However, protein measurement techniques have historically been limited to measuring only a few proteins at a time. This review provides an overview of the evolution of analytical techniques for multiplex analysis of protein biomarkers developed to address these limitations. These significant technological advancements have increased the multiplexing capacity of protein measurement techniques, enabling the routine measurement of hundreds or thousands of proteins from a single sample in a single instrument run. We will also discuss the implications of a more nuanced understanding of diseases and their associated biomarkers brought about by these new techniques. Lastly, we will explore potential future directions and implications in this rapidly evolving field.

2. Protein biomarker

2.1. Development of protein biomarkers

An ideal biomarker should be disease-specific, sensitive to alterations in disease state, minimally invasive, cost-effective, and biologically relevant, even if its applicability is limited to a specific clinical context. Due to these demanding criteria, biomarker development includes a discovery phase, and a verification/validation phase[6,7]. In the discovery phase, comprehensive analysis identifies differentially expressed proteins between disease and normal states, but this phase typically involves a small study population and high costs. To address potential false discoveries from small sample sizes, candidates undergo target-driven quantitative analysis in the validation phase to confirm differential expression in larger, independent sample cohorts. The verification step accounts for broader population variations before candidates proceed to assay optimization and analytical and clinical validation. Ultimately, validated biomarker assays are commercialized to meet clinical standards.

2.2. Increasing protein biomarker dimensionality

Under the one-gene-one-protein hypothesis[8] the number of potential biomarkers could be inferred from the number of protein coding genes. As of 2023 this is estimated to be 19,370[9], although physical evidence for transcribed gene products only exists for ~90% of these protein coding genes[10]. However, the complexity of protein structures and functions extends beyond genomic sequences. Variants such as genetic mutations, alternative splicing events, and post-translational modifications (PTMs) contribute to diverse array of proteoforms derived from the same genes[11] (Figure 1).

Figure 1:

Technological advancements have increased protein biomarker dimensionality. Proteins can exist in a myriad of configurations including sequence isoforms, post-translational modifications (PTMs) or chemical adducts. Moreover, the multiplex protein measurement technologies used, and which ultimately influence biomarker dimensionality, have evolved over time from relative protein quantification between disease states, to absolute concentration of protein targets, spatial proteomics for a region of interest, and most recently classification based on single cell measurements. Created using Inkscape 1.3 by Tao Liu.

For example, the protein coding regions of a gene consist of exons and introns, sequences of nucleic acids that can be selectively ordered or transcribed to ultimately modify a protein’s final amino acid sequence[12]. These variations result in different isoforms derived from the same gene, complicating detection and analysis as a single gene can produce multiple proteins with diverse biological functions[13,14]. While research continues to elucidate the role of protein isoforms, they have already been noted in biomarker discovery such as when Safikhani et al. reported specific isoforms of IGF2BP2, NECTIN4, ITGB6, and KLHDC9 as biomarkers predicting response for cytotoxic and targeted cancer drug treatments.[15]

In addition to differences in primary amino acid sequences, proteins can be chemically modified after translation either through specific PTMs or broader chemical adducts. There are over 400 unique PTMs influencing protein subcellular location, function, and degradation[16,17]. Unlike PTMs, chemical adducts are non-mediated chemical modifications of proteins. These can result from altered cellular states such as oxidative stress and/or exposure to anthropogenic chemicals[18,19]. With a growing understanding on the interplay in chemical exposure and human health[20,21], monitoring protein adducts could serve as a powerful biomarker for predicting diseases.

Beyond the biomarkers themselves, the design of the study used to identify them profoundly affects their dimensionality. Several critical considerations must be addressed: the establishment of an absolute concentration threshold for a protein to qualify as a biomarker, the adequacy of detecting the presence of a specific protein to fulfill the criteria of a biomarker, and whether expression within a particular cell type or tissue location is sufficient[22]. These considerations, alongside numerous other factors enabled by advancements in technology, must be meticulously evaluated in the context of protein biomarker discovery. The integration of these dimensions is crucial for developing a comprehensive understanding and accurate identification of potential biomarkers.

Combinatorically, all the above results in an exponential growth in the number of possible protein biomarkers. Biomarkers require rigorous evaluation across analytical, clinical, and practical domains to ensure their reliability and utility. Analytical criteria focus on specificity, sensitivity, precision, accuracy, and robustness. Clinical criteria emphasize predictive value, consistency, and relevance to disease progression. Practical feasibility includes cost-effectiveness, rapid turnaround, and ease of use. Compliance with regulatory and ethical standards is critical, alongside integration with multi-omics data and standardized protocols for reproducibility. While some limitations, such as protein copy numbers in cells, curb the number of possible protein isoforms, technological advances with their promises of lower detection limits and more comprehensive coverage, many of which are described below, will invariably necessitate a more granular definition of protein biomarkers to differentiate functional isoforms and PTMs from biological noise[16] for the accurate prediction and identification of disease.

3. Detection of protein biomarkers

Proteins, unlike genetic material, cannot be artificially amplified (e.g., via polymerase-chain reaction, PCR). Coupled with frequent limitations on sample input, a dynamic range spanning several orders of magnitude[23], expression varying across cell and tissue types[24], and differing physicochemical properties between protein species, advances in multiplexed protein measurements has often lagged behind the genetic counterparts. In response, the development of analytical techniques for multiplexed analysis of protein biomarkers has diverged into two broad categories: affinity-based and mass spectrometry (MS)-based technologies. For the former, interactions between a specific binding agent and a target protein forms the basis of detection, whereas in the latter, direct measurements for intact proteins or proteolytic peptides via MS are used. A survey of both technological streams, and their evolution in multiplexed capacity is discussed below.

3.1. Affinity-based technology

Affinity proteomics is the field of proteome analysis based on the use of antibodies and other binding reagents as protein-specific detection probes. The first of these techniques, western-blots and enzyme-linked immunosorbent assays (ELISA), were developed in the 1970s, and unlike previously available techniques, enabled specific protein measurements[25,26] (Figure 2). Western blot and ELISA techniques can be easily adapted for different target proteins by simply changing the antibodies used for capture and detection in the assays. The availability of existing antibodies, coupled with methodological refinements, has allowed these techniques to underpin protein biomarker studies for decades[27].

Figure 2:

Schematic overview of selected affinity and LC-MS based multiplexed proteomics analysis techniques. While most immunoaffinity approaches are dependent on the manufacture, LC-MS based proteomics are readily adaptable to suite an individual research question. Created using Inkscape 1.3 by Tao Liu.

However, as demand for more complex and comprehensive proteomics data increases, the limited multiplexing capacity and throughput of these assays have left researchers wanting. The considerations of these assays are not primarily due to the antibody-protein interactions, but rather to the challenges in detecting these binding events. Typically, antibody-protein interactions are monitored using a secondary detection antibody conjugated to an enzyme that catalyzes a colorimetric or luminescence reaction, which is then detected spectroscopically. Even when extending to fluorescent dyes for detection, there is a limit to how many different dyes can be used simultaneously without significant spectral overlap which can produce false positive signals or reduce sensitivity. Consequently, affinity-based proteomics has pivoted to flow cytometry and microfluidics to enhance multiplexed measurements while minimizing the requirement for higher sample input [28]. Several multiplexed affinity-based technologies have been developed, such as Luminex, immuno-PCR, proximity ligation assay (PLA), proximity extension assay (PEA), and slow off-rate modified aptamer assay (SOMAmer). Table 1 provides a brief survey of the technologies discussed below.

Table 1:

Overview of Prominent affinity-based technologies for multiplexed protein biomarker analysis.

Feature	ELISA	xMAP	PEA	SomaScan
Primary Vendor	Multiple vendors	Luminex	O-link	SomaLogic
Multiplexing	Low (single analyte)	Moderate (≤ 500 analytes)	High (5400 + for Olink Explore HT)	High (11,000 proteins)
Sample Volume	50-100 μl/well	25-50 μl	1 - 6μl	2 × 55 μl (blood); 2 × 260 μl(urine)
Throughput	Low ~40 samples/plate	Moderate ~384samples/plate (With 7 standards for each analyte)	High 88samples/plate	High 85 samples /plate
Detection Method	Colorimetric or fluorescent readout	Fluorescent flow cytometry	qPCR, NGS amplification of DNA tag	Fluorescent microarray readout of aptamer binding
Dynamic range	∼2 logs	4–5 logs	10 logs	10 logs
Sensitivity	pg/mL	pg/mL	pg/mL~fg/mL	pg/mL~fg/mL
Cost	Low cost per single analyte	Moderate (per assay)	High (per panel, but lower per protein assay in panels)	High (per panel, but lower per protein assay in panels)

3.1.1. Multianalyte profiling (xMAP)

Luminex, established in the late 90’s, was one of the first affinity-based analytical techniques for multiplexed protein measurements to achieve commercial success due to their innovative multianalyte profiling (xMAP) technology[29]. Herein a target-specific antibodies are conjugated to polystyrene microspheres, exposed to the sample. Detection is achieved using biotinylated secondary antibodies and a streptavidin-phycoerythrin (PE) reporter, forming a sandwich complex similar to the ELISA assay described earlier[30]. While each capture and secondary antibody are unique to the target analyte, the streptavidin-PE reporter means that only one “green” laser is needed for detection/quantification. The breakthrough feature of the xMAP platform lies in using distinct fluorophore dye ratios on the microspheres, allowing for a second “red” laser to differentiate bead colors and thereby identify each analyte uniquely allowing only one analyte to be detected per bead region [30]. When coupled together into a dual-laser flow cytometry setup, the xMAP technology can detect up to 500 distinct analytes from ≥25 μl serum input. This can be used to screen soluble and secreted proteins, allowing for detection of cytokines, chemokines and autoimmune antibody with high sensitivity and a low detection limit of 10 pg/ml [31]. The versatility of xMAP technology has been demonstrated in various studies. For instance, screening 17 soluble immune checkpoint proteins in serum from healthy donors and pancreatic ductal adenocarcinoma patients showed that high sCD40 levels are associated with poor prognosis[32].

In addition, this technology continues to innovate with the introduction of a second violet laser (Intelliflex dual-laser platform) for reporter channel, enabling simultaneous detection of two analytes. This enhancement ensures data collection efficiency and reduces analysis time in various clinical and research applications. For example, it facilitates multiplexed analysis by measuring 2 different isotypes per antigen within a single assay setup. Multiple SARS-CoV-2 spike-specific IgG subclasses[33] and cytokines[34] were measured. Compared with ELISA, the ability to measure multiple antibody responses simultaneously provides significant advantages over traditional single-analyte assays, enhancing the utility for diagnostic, serosurveillance, and vaccine efficacy studies[35] Despite certain limitations, the established presence of xMAP technology in clinical diagnostics ensures its continued relevance in biomarker research.

3.1.2. Immuno-PCR (IPCR)

First developed in the 1990s, immuno-PCR (IPCR) replaces the enzyme-linked antibody in ELISA with an antibody-DNA conjugate. The DNA attached to the antibody serves as a template for PCR amplification, combining the specificity of antibody-based detection with the signal amplification power of PCR[36,37]. For example, IPCR can increase the sensitivity of PSA detection by nearly 1000-fold compared to ELISA, achieving a limit of detection (LOD) of 0.1 pM [38]. Additionally, an assay developed enhancing detection by 16,000-fold for IL-18 in patients with immunodeficiency[39]. Despite its superior sensitivity, increased cross interactions among antibodies present a challenge for multiplexing, as they can introduce false positive signals. Consequently, studies have conservatively limited the number of target proteins to 10 or fewer[40]. Lastly, IPCR requires careful optimization of both immunoassay and PCR components to achieve reliable results[41].

3.1.3. Proximity ligation/extension assay (PLA/PEA)

Expanding from IPCR and the combination of immunoassays and PCR, is the development of proximity ligation assays (PLA) and proximity extension assays (PEA). A significant development in these assays is the utilization of dual antibody recognition of targets, with each antibody linked to DNA oligonucleotides.

In PLA, the two DNA strands are complementary to a third connector DNA oligonucleotide. When in proximity, i.e., both antibodies are bound to the same target protein, all three strands can hybridize, then DNA ligase is used to create a circular DNA molecule, subsequently serving as a template for rolling circle amplification or quantitative PCR (qPCR)[42]. This technology has been applied to detect more than 20 and 74 biomarkers in clinical plasma samples using panels of 7-plex and 24-plex PLA reactions [43–45]. This innovation in NULISA, a proteomic liquid biopsy platform, enhances traditional proximity ligation assays by ~10,000-fold to achieve attomolar sensitivity[46]. It enables highly multiplexed quantification of proteins with a fully automated workflow, improving the sensitivity and efficiency of proteomic analyses for research and diagnostic applications. In addition, PLA has found use in spatial measurements, as discussed below, as amplified DNA strands intercalated with fluorescent dyes are visualized as distinctive spots [46,47].

In PEA the two antibody DNA oligonucleotides are complementary to each other. Thus, when in proximity they partially hybridize to each other, DNA polymerase completes the hybridization which is subsequently amplified and monitored via qPCR [45] (Figure 2). Like the previous immunoaffinity techniques discussed above, its multiplexing capacity is limited due to spectral overlap in the spectroscopically based detection methods. To address this limitation, samples are diverted using microfluidics into multiplex wells, each containing a subset of assay panel antibodies. This ensures that each well remains within the multiplexed detection capability of qPCR, while the use of microfluidics and the signal amplification provided by qPCR eliminates the need for additional sample input. The PEA assay has seen commercially success under the name Olink. Olink PEA Target utilizes a qPCR-based readout; measurement is typically reported as relative concentration units (NPX values). However, a few new Olink Target panels were recently developed (eg. Target 48 or Flex), these new panels are low-plex (5 to 45 proteins) and include standard curves and calibrators, providing actual protein concentrations (pg/mL). Olink Explore uses NGS (Next-Generation Sequencing) to significantly increase multiplexing capacity, detecting over 5,400 proteins, as relative concentrations. The Olink Explore assays are validated for analytical performance across a quantifiable range, with lower and upper limits of quantification averaging a 500-fold difference in antigen concentration [47]. Validation and robustness tests assess sensitivity, dynamic range, specificity, precision, endogenous interference, and detectability in both healthy and disease samples. Dynamic range and precision are determined using a dilution series of recombinant antigen, defining LOD as three standard deviations above the negative control, and ensuring %CV and accuracy below 30%. The quantifiable range averages 2.7 log10, with intra- and inter-assay precision of 7.8% and 10.6%, respectively. Detectability criteria require assays to detect at least 50% of samples, and dual antibody recognition ensures 99.8% specificity, reducing non-specific binding. ExAmp chemistry in NGS protocols addresses index hopping, ensuring accurate sequencing results. More recently, the Olink assay has demonstrated that it can simultaneously quantify over 3,000 proteins in a sample with microliter amount in over 54,000 participant samples from the UK Biobank (UKB)[44,45]. In addition, a longitudinal study of Alzheimer’s disease showed intra-, and inter-panel CVs were consistently below 15% for most proteins[50]. The study demonstrated that intra-plate CVs strongly correlated with NPX values, indicating that higher NPX values generally correspond to lower CVs. For inter-panel comparisons, they demonstrated high correlations, suggesting consistency in measurements between panels, the correlation of values obtained from the same protein on different panels remained high, ranging from 0.73 to 0.97. However, absolute NPX values may differ due to variable dilution factors between the panels, therefore while relative concentrations are consistent, direct comparisons of exact NPX values should be interpreted carefully.

3.1.4. Slow off-rate modified aptamer assay (SOMAmer)

While antibodies, with their specificity, formed the primary detection tool in the previous assays, they also present a major challenge in assay development, as antibody development is a lengthy process involving immunization, hybridoma, screening, testing, and selection[51].

In response Gold et al. introduced Slow Off-rate Modified Aptamers (SOMAmers)[52]. Based on aptamers, short single-stranded oligonucleotides that can bind specifically to target molecules with high affinity, SOMAmers incorporate additional polymerase compatible synthetic nucleic acids and are manufactured abiotically and developed using Systematic Evolution of Ligands by EXponential enrichment (SELEX)[52]. In essence, a large library of randomized single-stranded DNA sequences is incubated with a purified protein target; unbound sequences are removed, and bound sequences are subjected to amplification and this enriched subgroup is resubjected to binding. After several iterations, the final pool is sequenced, synthesized, and individually tested for assay compatibility. Aptamers, and SOMAmers in particular, are touted as being more selective than antibodies, and owing to their entirely synthetic manufacturing process, more reliably manufactured and reproducible[53,54].

To be incorporated into a detection platform, SOMAmers are additionally modified with a fluorophore, cleavable linker, and biotin moiety[52] (Figure 2). The SOMAmer binds to and becomes immobilized, on a streptavidin bead. When exposed to a biofluid sample, immobilized SOMAmers bind with their target proteins based on the individual SOMAmer design. Once proteins are bound, they are in turn biotinylated. Rinsing removes non-specific interactions, and UV light is used to break the linker, releasing the SOMAmer-protein complexes from the streptavidin beads. The complexes are then exposed to fresh beads, this time with the biotinylated proteins being immobilized. Denaturing of the proteins releases the SOMAmer reagents which then hybridize with complementary strands on a microarray chip. The fluorescence intensity detected on the microarray chip is proportional to the available epitopes in the original sample. The SomaScan platform can quantify up to 7,000 proteins in a single test and is reported to have dynamic range capabilities spanning eight orders of magnitude [55]. This broad dynamic range means that within a single panel, it can quantify proteins from low- to high-abundance, providing comprehensive coverage of proteome. The platform offers great sensitivity (lower detection limit of 40 fM) and ~5% median coefficients of variance (CV). Also, even with processing delays and within archived samples the platform offers high reproducibility and stability for protein measurements[56,57]. With this, the SomaScan platform was used to measure 4,907 proteins from 35,559 plasma and suggested potential novel protein quantitative trait loci (pQTLs) related with disease[58]. These advantages make the SomaScan platform a powerful tool for high-throughput proteomic analysis, enabling novel insights into disease.

3.2. Mass-spectrometry-based technology

In contrast to the affinity-based approaches discussed above, which infer analyte detection from measurements of a secondary property or response that correlates with the presence or concentration of the analyte, MS directly measures the analyte of interest. MS works by ionizing chemical compounds to generate charged molecules (or molecule fragments) and then measuring their mass-to-charge ratios (m/z). Since mass is an intrinsic property of all molecules, MS is considered a universal measurement technique. Consequently, MS-based approaches are readily adaptable to individual research questions such as identifying PTMs or identifying new protein isomers, all without the need for developing specific antibodies or aptamers.

In protein biomarker research, MS-based approaches often employ ‘bottom-up’ strategies, wherein proteins of interest are enzymatically digested into peptides, which are then sequenced using MS and matched to known proteins[59]. A typical MS experiment can be broadly divided into three sections: sample preparation, MS analysis, and data analysis. Experiments begin with the enrichment of proteins from biological samples. Proteins are then digested into peptides, most commonly using the protease trypsin. Prior to MS analysis, peptides are separated via liquid chromatography (LC) and ionized through electrospray as they are injected into the instrument. Within the MS, peptides are fragmented to generate spectra that are instrumental for peptide and protein identification in subsequent data analysis steps (Figure 2).

For a more comprehensive introduction to MS-based proteomics, readers are directed to several reviews and tutorials[59–61]. Due to the interrelated nature of MS-based proteomics, innovations at all steps of this process—including sample preparation, sample measurement, and data analysis—have led to impressive increases in the multiplexing capacity of MS-based proteomics. Each of these aspects is discussed in turn below.

3.2.1. MS sample preparation

The culmination of MS-based proteomics is the measurement of the m/z of ions to identify and quantify molecules in samples. In bottom-up proteomics, this involves peptides derived from the digestion of protein-containing samples. MS-based proteomics does not require proteins to be in their native state, and thus any protein-containing tissue or biofluid can be analyzed. Due to the prominence of biofluids (i.e., plasma and serum) in biomarker research, there is a corresponding emphasis on biofluid sample preparation for proteomics[62]. The principal obstacle for biofluid analysis is the enormous concentration differences between high-abundance and low-abundance proteins, spanning over 10 orders of magnitude, which makes it challenging to detect and quantify less abundant proteins that may be potential biomarkers[63].

While researchers may opt to investigate ‘neat’ plasma or serum[64], the dynamic range issue limits the number of identifiable features. Highly abundant proteins, such as albumin and immunoglobulins, are often depleted using affinity-based columns from samples as an initial step in sample preparation[65]. Additional affinity-based depletion approaches can include commercially available spin-columns and antibodies[66]. A recently developed high-throughput depletion strategy utilizes perchloric acid, which generally precipitates most proteins regardless of their abundance. Both high-abundance and low-abundance proteins can be precipitated. This approach was used by Viode et al. to achieve a 60% (3,198 vs. 1,919 proteins) increase in the number of protein identifications compared to a traditional sample preparation in a study analyzing more than 3,000 plasma samples[67] As small peptides or protein fragments tend to precipitate less, the use of perchloric acid may increase detection of peptides with non-specific cleavage[68].

A complementary approach to depletion is to enrich for low-abundance proteins. The most promising of these approaches utilizes nanoparticles (NPs), engineered with different physicochemical properties to induce the formation of protein coronas (i.e., layers)[69]. The composition of proteins in the corona effectively compresses the proteomics dynamic range and afford unbiased sampling of proteins across a wide abundance spectrum[70,71]. Despite its novelty, the SEER platform, one of the most successful ventures in this space, has already been used to survey 2,513 cases and control subjects as part of a multi-omics profiling for the blood-based early detection of lung cancer[72]. Altogether, these features make CERES’ hydrogel nanoparticles readily amenable to automation and suitable for inclusion in future clinical studies[73]. Hydrogel nanoparticles containing a core-shell pores allow concentrate low molecular weight proteins[74,75]. CERES Nanosciences has made significant strides in the utilization of hydrogel nanoparticles for protein capture. The unique porosity and binding properties of these nanoparticles enable the efficient sequestration and concentration of low-abundance proteins, facilitating biomarker discovery and enhancing diagnostic capabilities[76,77]. High-resolution isoelectric focusing (HiRIEF) for the separation and detection of proteins based on their isoelectric points (pI) also enhances high resolution and sensitivity, enabling the identification of low-abundance proteins and providing deep proteome coverage[78,79].

A promising strategy for future development involves targeting extracellular vesicles (EVs), small membrane-bound particles that transport proteins and other macromolecules between cells[80]. Effective methods for isolating EVs include differential centrifugation, density gradient centrifugation, size-exclusion chromatography, and immunoaffinity capture[81]. For example, Mag-Net uses a combination of surface chemistry and particle porosity to capture and enrich EV-like proteins from biofluids while simultaneously depleting abundant plasma and lipoproteins [82]. This approach holds promise for future developments in biomarker discovery and disease diagnosis, as it allows for the in-depth study of the extracellular vesicle proteome. These advancements underscore the ongoing development of methods to detect low-abundance proteins, further enhancing the capabilities of proteomic research. Altogether, considering extracellular vesicles and relationships between circulating proteins, as well as implementing novel sampling techniques, will advance our understanding of the dynamics of human health and disease[83].

Another consideration in MS-based proteomics sample preparation is the downstream analysis strategy. The simplest approach is to analyze samples without isotopic labeling (i.e., label-free), but isotopic labeling techniques can significantly enhance multiplexing capacity to achieve higher throughput. The highest-throughput method employs tandem mass tags (TMT)[84]. In this approach, peptides from different samples are labeled with isotopically distinct isobaric tags[85]. These tags consist of a reporter group, a balance group, and an amine-reactive group. Each sample is labeled with a different TMT reagent, which reacts with peptide N-termini and lysine side chains. Labeled peptides from different samples are then combined into a single mixture and analyzed in a single LC-MS/MS run. During MS/MS analysis, the labeled peptides share the same parent ion mass but release isotopically distinct reporter ions upon fragmentation. The relative abundance of peptides is determined by comparing the intensities of these reporter ions[86]. The multiplexing capacity, the number of samples that can be quantified in a single LC-MS run, has increased from 6-plex to 10-, 18-, and most recently up to 35-plex[86–88]. While the additional complexity incurred from TMT-prepared samples requires careful considerations in sample-preparation and experimental conditions, it has remained a stable technology in MS-based protein biomarker studies[85,89]. Moreover, the nuances of TMT analysis lends it favorably for the study of PTMs and low-input samples through the use of a ‘carrier’ channel that contains a large amount of protein relative to the experimental channels with smaller amounts, boosting the signal strength to improve peptide identification and quantification[90].

3.2.2. MS data acquisition and analysis

Like sample preparation, MS setup and experiment design will greatly impact what and how a study can measure peptides for biomarker discovery and validation. MS data acquisition can be broadly categorized into either targeted, data-dependent acquisition (DDA), or data-independent acquisition (DIA) strategies (Figure 2, Table 2). Targeted approaches such as selected/multiple or parallel reaction monitoring (SRM/MRM or PRM) are targeted mass spectrometry techniques used for the detection and quantification of specific peptides[91]. SRM/MRM is performed on triple quadrupole instruments to simultaneously monitor multiple pairs of specific precursor and product ions (i.e., transitions). PRM, performed on hybrid MS instruments (e.g., Q-Orbitrap, Q-TOF), provides high-resolution detection of all product ions. Despite limited multiplexing capacity, targeted approaches offer increased sensitivity and quantitative accuracy[92]. Moreover, targeted MS approach provides a straightforward path for the application of MS in clinical settings. Currently, there are multiple large-scale studies that develop methods and identify protein biomarker panels for different diseases and furthermore assembled protein panels in targeted MS can have different utilities such as the diagnosis[93], distinction between benign and malignant[94], and prediction of recurrence[95] in prostate cancer.

Table 2.

Overview of Mass spectrometry data acquisition methods for protein biomarker analysis

Feature	DDA	DIA	PRM	SRM, MRM
Application	Discovery	Discovery, Verification	Verification	Verification, Validation
Detection method	Selects the most intense peptide ions for fragmentation	Fragments all peptide ions within defined mass windows	Fragments preselected peptide ions	Fragments preselected peptide ions
Depth of coverage	High	High	Moderate	Low to moderate
Sample throughput	Low to moderate	High	Moderate	High
Sensitivity	Moderate	High	High	Very high
Reproducibility	Moderate	High	High	Very high

DDA and DIA are more often employed in discovery-based approaches or hypothesis-free investigations. In DDA a subset of the most abundant ions (i.e., Top-N ions) are sequentially selected for fragmentation. DDA data analysis involves identifying peptide-spectrum matches (PSMs) by comparing observed spectra against a database of theoretical peptide spectra. DDA is obligatory for TMT-based studies and is the preferred approach for investigating novel PTMs due to the simpler mass spectra[96]. In DIA, preselected windows of ions are systematically fragmented and all precursor ions within predefined m/z ranges are analyzed. DIA data analysis involves deconvoluting complex spectra to quantify and identify peptides by using comprehensive spectral libraries and sophisticated computational algorithms[97,98]. DIA has emerged as the predominant approach to label-free analysis as it enables more comprehensive proteome coverage in shorter analysis times[99].

A comprehensive comparison of MS based data acquisition and analysis approaches is beyond the scope of this paper. However, as discussed below, advances in MS-based sample preparation, instrumentation, and data analysis pipelines have extensively improved the sensitivity, reproducibility, and coverage universally across MS acquisition methods. Moreover, the latest models of mass spectrometers offer enhanced resolution, accuracy and sensitivity. The recently introduced asymmetric track lossless (Astral) analyzer (Thermo Scientific), employing a novel ion-trapping mechanism, has pushed the limit of DIA to narrow windows down to a DDA-like 2 m/z, achieving high coverage as well as reproducibility [100,101]. Application of Astral to single-cell proteomics has resulted in impressive coverage of over 5,000 proteins per cell at high throughput, which offers new possibilities for large-scale biomarker analysis [102,103]. In parallel, advances in the timsTOF (trapped ion mobility spectrometry-time-of-flight) MS series (Bruker), allowing the separation of ions based on their mobility in gas phase and incorporating Parallel Accumulation Serial Fragmentation (PASEF) in diaPASEF mode, maximize ion usage by accumulating ions in parallel and then fragmenting them serially [104], has demonstrated a significant boost in sensitivity [105,106]. Additionally, to address the challenges in targeting thousands of peptides, Plubell et al. employed a Stellar MS (Thermo Scientific), a novel hybrid high-speed mass spectrometer that combines the robustness of triple quadrupoles with the enhanced capabilities of an advanced linear ion trap analyzer to allow real-time retention time alignment [54] . Integrating advances in sample preparation with fast and sensitive MS acquisition could also be crucial for large-scale assay development using small clinical samples. The increased sensitivity of these instruments enables researchers to detect and quantify low-abundance proteins and other biomolecules with greater confidence. This capability is particularly valuable in clinical research, where the detection of biomarkers at low concentrations can be crucial for early diagnosis and treatment monitoring [107].

3.3. Comparison of affinity- and MS-based technologies

The push towards more comprehensive proteomics coverage has centered Olink, SomaScan, and MS-based proteomics as the front-runners for multiplexed biomarker discovery. Despite each platform ostensibly claiming to measure many of the same soluble proteins, the mediocre correlation in measured protein quantities is telling, highlighting potential differences in sensitivity, specificity, and dynamic range of these technologies.

For example, the Atherosclerosis Risk in Communities (ARIC) study compared the platforms (SomaScan V4, 4,955 proteins; Olink Proseek V5003, 460 proteins) with 427 participants. They report: “approximately 1/2 of the overlapping proteins from the different platforms are highly (19%) or moderately (34%) correlated, but the other 1/2 approximately had poor correlations (r < 0.5) and may be capturing different information”[108]. They discussed differences in data reporting, sample storage, and overall assay conditions as possible confounding factors. A similar comparison of the SomaScan 11k and Olink Explore HT (4,443 overlapping assay targets) utilizing split plasma samples from 102 ARIC study participants reported a bimodal distribution in correlation between targets[109]. This comparison revealed that higher abundance proteins had a greater correlation than lower- abundance proteins, which often fell below the assay’s reported LOD. Differences in assay quantification and reporting were noted as explanations for this.

In addition, Affinity-based proteomics using the aptamer-based SomaScan v4 assay and antibody-based Olink assays analyzed 871 common protein targets across over 10,000 participants. This comparison revealed that about 36.3% of protein-phenotype connections were unique to one platform, showing distinct aspects of protein biology. Median correlation coefficients for overlapping measurements were found to vary widely, influenced by technical factors and genetic variants. The study highlighted significant genotype effects on correlations between platforms, such as with YKL-40 and IL-7 receptor subunit alpha. Integrating these proteomic techniques with genetic data enhances understanding of disease mechanisms and drug targets, setting a benchmark for future research[110].

Another study investigating protein quantitative trait locus (pQTL) compared data from over 50,000 UKB participants using Olink Explore 3072 to SomaScan V4 data from 36,000 plasmas. They reported a high degree of internal consistency between assays but similar discordance in measurements for many lower-abundant protein targets across the two panels[111]. The authors discussed the possibility of sequence variants among explanations for this discrepancy. Aptamers and antibodies can interact with non-target macromolecules, exhibit different binding affinities to isoforms of the same target protein, and be influenced by a myriad of experimental factors. This cross-reactivity, combined with the fact that a single amino acid change in the binding site can significantly alter binding affinity, underscores the importance of careful validation to minimize off-target interactions[112].

MS-based proteomics, unaffected by factors influencing binding affinity, provides extensive profiling of proteins and enables the simultaneous identification and quantification of thousands of proteins. A comparison of SomaScan, Olink, and the MS-based Seer Proteograph platform suggests that up to a third of affinity pQTLs may be affected by epitope effects[73]. Despite the promise of MS to rival the affinity-based techniques, track PTMs, and analyze non-soluble proteins, MS-based methods are hampered by the need for extensive sample preparation and sophisticated instrumentation. In the face of a panoply of sample preparation and data analysis methods, clinicians can hardly be blamed for choosing the one-stop-shop convenience afforded by Olink and SomaScan. It should be noted that several vendors have begun to offer contract research services in response to these criticisms.

Another approach has been to combine these techniques for comprehensive coverage. For example, Dammer et al. developed a tunable median polish of ratio (TAMPOR) approach for batch correction and harmonization of MS, Olink, and SomaScan data[113]. The algorithm adjusts the central tendency of protein abundances by calculating ratios within each batch and then log-transforming these ratios. These log-transformed values are centered to a median value of zero for each sample and then adjusted back to their original scale. The process iteratively repeats until batch effects are minimized and the data’s central tendency is harmonized. Applying this approach to actual clinical samples, TAMPOR helped integrate data from multiple cohorts and platforms (TMT-MS and SomaScan), revealing a robust and comprehensive protein co-expression network related to Alzheimer’s disease in cerebrospinal fluids[114]. This approach facilitated the discovery of novel protein alterations and modules that would be missed if only a single platform was used, such as those related to APOE ε4 and atomoxetine treatment.

The choice among quantitative platforms should depend on specific research needs, proteins of interest, and the recognition that correlations between platforms for overlapping proteins may be inconsistent, necessitating careful interpretation of results. As each platform’s unique strengths highlight the benefits of using complementary approaches to advance biomarker research, integrating these platforms enhances the reliability and breadth of detectable biomarkers.

4. Future directions

The landscape of biomarker discovery and disease mechanism elucidation has been revolutionized by advancements in multi-omics, single-cell and spatial proteomics, and high-throughput screening technologies. These cutting-edge approaches allow researchers to comprehensively analyze and integrate diverse layers of biological data, providing unprecedented insights into the molecular underpinnings of complex diseases. By leveraging these innovative techniques, researchers are now poised to uncover novel biomarkers, understand intricate cellular heterogeneity, and enhance the efficiency and scalability of proteomic analyses. The following sections delve into these transformative methodologies and their significant impact on biomedical research.

4.1. Rectangular search strategies

The traditional ‘triangular’ biomarker search strategy, characterized by a low sample size with a high number of potential protein candidates followed by validation with a large sample size and targeted analysis of a few protein candidates, is poised for upheaval due to advances in multiplexing capacity. The ‘rectangular’ strategy entails measuring hundreds or thousands of proteins across a large number of subjects in two or more different cohorts[107,115]. This approach yields much greater statistical power and, by analyzing both discovery and validation cohorts simultaneously using consistent, robust proteomic techniques, minimizes the impact of study-specific confounders and enhances the identification of reliable biomarkers. For MS-based studies, including biomarker criteria adaptable for targeted assay development can further streamline the transition from validation to clinical use[116]. For example, a study across 15 hospitals in the United States analyzed 2,934 plasma samples from 1,117 hospitalized COVID-19 patients, identifying 2,910 proteins[85]. This analysis revealed that concentrations of various biomarkers, such as those related to neutrophil extracellular traps (NETs) and heart damage markers, were associated with fatal outcomes. By concurrently evaluating discovery and replication cohorts, this study demonstrated a robust approach to identifying reliable prognostic biomarkers and validated their findings with large-scale statistical power, minimizing study-specific confounders. In another example, a comprehensive analysis of 2,147 plasma samples from children and adolescents aged 5-20 years, along with an additional 1,558 samples from replication cohorts, quantified 1,216 proteins[117]. This study showed that genetic variants, age, sex, and BMI significantly influence plasma proteome variation. By concurrently evaluating discovery and replication cohorts, the study showcased a robust approach to identifying genetic determinants, validated their findings with substantial statistical power, and minimized confounding factors, enhancing the reliability of identified biomarkers. Altogether, this approach to biomarker discovery is only economically feasible due to the previously discussed technological advances.

4.2. Multi-omics

Multi-omics approaches integrate different layer of omics data such as genome, proteome, metabolome, and lipidome to generate comprehensive insights into diseases mechanisms. This integration allows for discovery of novel molecular relationships among biological molecules, uncovering new networks that might be missed by single omics approaches. By combining multiple layers of omics data, researchers can identify more accurate and reliable biomarkers, facilitating the understanding of complex molecular mechanisms of diseases and aiding in the discovery of potential therapeutic targets and development of personalized treatment approaches.

For example, the Clinical Proteomic Tumor Analysis Consortium (CPTAC) has reported a range of tumor-tissue studies that integrate genomic and proteomics to identify novel molecular subtypes and therapeutic targets across various cancer types[118]. By integrating various omics data, researchers can identify biomarkers that are more robust and clinically relevant, such as driver genes and proteins critical for disease progression[119], and treatment response[120]. Expanding the focus from primarily studying phosphorylation to include other PTMs such as ubiquitination, acetylation[121], and glycosylation [122] helps to elucidate underlying signaling mechanisms. The crucial role of metabolic processes in cancer development has also been highlighted when incorporating metabolomics and lipidomics into analysis[123,124].

Beyond cancer research, multi-omics approaches have also proven valuable in other fields. For instance, longitudinal multi-omics study has revealed molecular changes with nonlinear patterns during aging, identifying strong correlations among metabolites, cytokines, and the oral microbiome with age, including major molecular shifts related to metabolic pathways[125]. Additionally, a study incorporating 18 diverse omics platforms elucidated significant molecular interactions relevant to diseases like diabetes, cardiovascular, and autoimmune disorders[126]. This study demonstrated platform-specific capabilities to explain phenotypic variations and uncovered novel biomarkers and potential therapeutic targets.

The integration of multi-omics data is a complex but pivotal task for gaining deeper insight into biological systems. Various computational tools and algorithms have been developed to facilitate the integration and analysis. Tools such as OmicsAnalyst [127] COSMOS [128], SignalingProfiler [129], and Multiomics2Targets [130] offer unique capabilities to address different aspects of multi-omic analysis, from data integration to visualization and drug target identification. As the fields evolves, it is essential to continuously monitor new tools and updates, leveraging the strengths of multiple platforms in a complementary manner to solve complex biological mechanisms.

Despite the significant progress, several challenges remain in the field of multi-omics. The major hurdle is the annotation of unknown molecules, which often limits the functional interpretation of omics data. Also, different experimental conditions and techniques may lead to variability and lack of reproducibility. Therefore, the standardization of data generation, processing, and analysis is crucial ensuring consistency and reproducibility across different studies. The development of more sophisticated computational tools and frameworks for data integration will be essential for harnessing the full potential of multi-omics data.

The future of multi-omics integration holds immense potential for advancing our understanding of complex biological systems and improving disease diagnosis and treatment. By leveraging data from various omics platforms, researchers can achieve a more comprehensive understanding of the interactions and networks that govern protein function. This system biology approach uncovers novel biomarker candidates that may not be apparent when studying proteins in isolation, ultimately enhancing the utility of biomarkers in clinical practice.

4.3. Single-cell and spatial proteomics

Proteome profiling of bulk tissue samples has yielded valuable biomarker insights but often overlooks the distinct characteristics of different cell types within tissues. This limitation is particularly problematic when studying diseases like cancer, where certain cell types can be critical despite representing only a small fraction of the overall sample. To address this, single-cell proteomics methods now allow researchers to identify cell-type-specific proteomes within a tissue. Recent advances in single-cell proteome analysis, which creates a single-cell suspension of tissue followed by cell sorting and analysis, have offered insights into the cell type composition of tissues. Single-cell proteome analysis technologies such as nanoPOTS (nanodroplet processing in one pot for trace samples)[131,132], SOP-MS (surfactant-assisted one-pot sample preparation coupled with mass spectrometry)[133], and SCoPE-MS (Single Cell ProtEomics by Mass Spectrometry)[90] have made significant progress in MS-based field and show potential for clinical applications.

Compared to scRNA-seq, proteomics generally has lower sensitivity and struggles with detecting low-abundance proteins[134]. It also typically analyzes fewer (e.g., >100x less) cells due to limitations in sensitivity and throughput of current mass spectrometry techniques. However, proteomics offers functional insights by identifying active proteins, their modifications, and interactions, directly linking to cellular functions and phenotypes[135]. These aspects cannot be achieved with transcriptomics alone. Deep proteomic profiling is crucial to distinguish and understand the distinct cellular contributions in each tissue type.

Spatial proteomics techniques retain the spatial context, enabling researchers to observe not only the presence of proteins but also their precise locations within tissue. This spatial mapping is especially valuable in cancer research, as it can identify regional variations in protein expression, as well as PTMs and cellular interactions that bulk analyses cannot capture[106]. For example, spatial proteomics has been used to define tumor boundaries more accurately, identify molecular markers that indicate invasive regions[136], and discover protein patterns associated with patient survival[137]. Such insights are crucial for developing more targeted therapies. The Human Protein Atlas (HPA) resource focuses on protein expression profiles in human tissues using conventional and multiplex immunohistochemistry, iteratively staining hundreds of tissue slices[138]. This approach enables the detailed spatial localization of proteins within tissues, including the identification of cell-specific protein expression. In contrast, sequencing-based spatial proteomics with mass spectrometry involves deep profiling of thousands of proteins from the same tissue voxel (pixel) derived from one tissue slice. Several techniques such as nanoPOTS[139], microPOTS (microdroplet processing in one pot for trace samples)[140], DVP (deep visual proteomics)[106,141] and wcSOP (wet collection of single tissue voxel and Surfactant-assisted One-Pot voxel processing)[142] have been recently reported to process minute tissue samples from laser-capture microdissection (LCM) to precisely isolate and analyze specific regions within a sample, uncovering localized proteome changes linked to disease progression. Wider application of the techniques to clinical samples may pave a way to elucidate the high heterogeneity of tumor tissue towards a better biomarker or therapeutic target.

Multiplexed imaging technologies, such as Imaging Mass Cytometry (IMC) and Multiplexed Ion Beam Imaging (MIBI), have furthered spatial proteomics by using metal-tagged antibodies to label multiple proteins within the same tissue section[143–145]. These methods provide high-resolution images showing not only the distribution of specific proteins but also the organization of different cell types within tissue microenvironments. For example, IMC has been used to study the tumor-immune microenvironment in melanoma, revealing unique immune cell arrangements associated with responses to immunotherapy[146]. Similarly, MIBI has offered insights into tuberculosis, showing how immune cell structures are organized to create localized immune responses within granulomas[147]. Although these techniques present challenges in terms of signal resolution and panel design, they are powerful tools for understanding complex tissue structures and functional niches[148]. In addition, recent advances in MALDI-IHC (MALDI mass spectrometry imaging combined with immunohistochemistry) offer a valuable approach for multiplexed, targeted and untargeted spatial proteomics[149]. Utilizing photocleavable mass tags, MALDI-IHC provides highly specific, spatially resolved protein imaging and complements untargeted bottom-up spatial proteomics, as demonstrated in breast cancer tissue analysis.

Spatial proteomics has also enhanced with multiplexed fluorescence-based methods, which use tagged antibodies to visualize multiple proteins within tissues, capturing complex interactions among cell types[150]. Studies of intestinal tissues, for example, have revealed how epithelial cells interact with immune cells in specific niches[151], while research on COVID-19 and vaccine responses has identified markers in lymph nodes that indicate different immune responses[152].

Altogether, spatial proteomics techniques have allowed researchers to obtain detailed spatial and temporal information about protein expression and localization. Providing information that is crucial for understanding the complex heterogeneity within tissues. This information holds enormous potential for the discovery of spatial and cell type-associated biomarkers, ultimately aiding in clinical applications and personalized medicine.

4.4. High-throughput screening

Biomarker development fundamentally depends on the number (n) of samples, identified proteins, and investigations. Across the board, the larger the n the more powerful the study[153]. However, economic realities force researchers into tradeoffs, often with larger study populations focused on a smaller subset of markers or vice versa. The advances in analytical techniques described above have sought to disrupt these tradeoffs by increasing the parallelization of protein measurements. All else being equal, measuring more proteins per unit of time can only improve a study. However, other approaches to the n problem have evolved alongside multiplexed analysis including parallelization in sample preparation, increased throughput for sample preparation and analysis, and a growing trend towards automation in all aspect of the analysis pipeline. Owing to synergistic effects between these elements they are often implemented simultaneously to increase n in all dimensions.

Whereas earlier efforts to marry these elements required institute-scale support, modern researchers can leverage relatively inexpensive and off-the-shelf components to readily achieve high-throughput analysis. Particle-based sample preparation is readily amenable to well plates, increasing sample preparation throughput [70,82,154]. Moreover, these workflows are readily amenable to automation on benchtop liquid handling robotics [155,156]. Whereas analysis was the typical bottleneck in biomarker studies, technological advances have dramatically increased throughput. Traditional (nano)liquid chromatography systems running a dozen of samples a day are being superseded by high-throughput systems such as the Evosep One, which can analyze upwards of 200 samples-per-day [157]. In addition, Ion Mobility Spectrometry (IMS) offers a potential alternative to Liquid Chromatography (LC) by separating ions based on their shape, size, and charge, rather than relying on their interaction with a stationary phase [158–160]. IMS can achieve rapid separation on a millisecond timescale, resulting in higher throughput and reduced sample analysis time. This technique can provide orthogonal separation, enhancing the resolution and sensitivity for complex mixtures. Additionally, IMS can separate isobaric species that are indistinguishable by LC, offering unique insights into the structural and conformational properties of biomolecules[161]. Indeed, with developments in IMS, the need for traditional chromatography is being reduced [158,162]. IMS allows the analysis of thousands of samples efficiently without the lengthy and complex steps typically required in chromatography.

Pressured by the demands for increased throughput in bioanalytical laboratories during the COVID-19 pandemic these technologies have already begun to be implemented in biomarker discovery[163,164]. Likewise, advancements in mass spectrometry technology, as detailed by Peters-Clarke et al. [165], have significantly enhanced the depth and breadth of proteome analysis. These developments include improvements in sensitivity, speed, and the ability to handle complex samples, allowing for more comprehensive and detailed proteome coverage than ever before.

In addition to the ever-increasing multiplexing capacity of both the OLink and SomaScan platforms, has been the development of even greater multiplexing capacity platforms. The Nautilus Proteome Analysis utilizes Protein Identification by Short-epitope Mapping (PrISM) [166,167]. This method employs affinity reagents that target short linear epitopes (short sequences of amino acids) on denatured proteins, which are proteins that have been unfolded to expose their linear amino acid sequences. These denatured proteins are then immobilized, or fixed in place, on a silicon chip. The development of ultra-dense ‘landing pads’ on a silicon chip results in a single protein per ‘landing pad’, thus detection of fluorescence detection of single-molecule events (i.e., one epitope per protein per pad), and sequential cycles, could provide single-molecule sensitivity across the entire proteome[166,168,169]. Similarly, the Platinum platform, in development by Quantum-SI, immobilizes peptides onto a semiconductor chip, and through a combination of fluorophore-labeled recognizers that reversibly bind to N-terminal amino acids and aminopeptidases can sequentially sequence peptides[170]. While both technologies promise comprehensive proteomic insights, they are still in early stages of development and has not yet been applied to comprehensive proteomic samples, let alone clinical biomarker studies.

As studies continue to expand in participant size and complexity, the efficiency gains from high-throughput systems are expected to meet new challenges. Consequently, future biomarker studies may find their primary limitations not in the speed of analysis, but in the acquisition of sufficient biological samples and the processing of increasingly complex data.

4.5. Data analysis strategy

The immense scope of modern and future clinical proteomic biomarker studies demands a considered approach to data storage, analysis, interpretation, and dissemination. Metadata is also crucial for experimental design and informs the data analysis approach from the beginning of projects. A thorough discussion of these is beyond the scope of this paper, but in essence there has been, and continues to be, an exponential growth in both the number of datasets and their respective sizes. For example, the PRIDE (PRoteomics IDEntifications) repository, a public data repository for sharing and storing MS-based proteomics data to facilitate data reuse and validation, reported ≈2,000 terabytes of data with no indication of a slowdown [171].

Despite the promise of hidden insights offered by reanalysis of data — such as elucidating the vast amounts of analytes detected by the MS that remain unidentified, the proteomics “dark matter” [13,172]— inconsistent and incomplete study metadata annotations can hinder reanalysis efforts and incorporation of multi-omics data for more thorough molecular profiling [173]. In the absence of proper metadata, valuable datasets may remain hidden and inaccessible to the broader research community. Metadata provides the necessary descriptors and keywords that enable efficient data discovery through search engines and data repositories. Ambitious efforts to consolidate disparate datasets from imaging, genomics, and proteomics are ongoing, but incorporation into routine studies has yet to be implemented[174].

Another consideration is scaling of the search and data-analysis algorithms to handle thousands or more samples, examples of recent implementation of these for massive parallel proteomics data analysis include quantms for parallel sample search[175], Scalable Match-Between-Runs (Scalable MBR) for cross-run normalization[176], and the PANOPLY cloud-based platform for proteogenomic data analysis[177].

From a researcher’s perspectives there are ongoing ethical considerations about the public disposition of increasingly granular and detailed clinical data with privacy and safety implications for cloud-based analysis[178]. Additionally, the cost of the infrastructure and operations of large-scale data analyses are not insignificant. Researchers leveraging cloud-based computation may avoid large up-front infrastructure cost but will incur ongoing operational cost; with the complexity of both approaches necessitating dedicated research personal to ensure efficient usage of increasingly costly resources. Lastly, while the promises of machine-learning/artificial intelligence are frequently espoused, these greatly amplify cost due to the vast datasets and computational power required to train these models[179]. Efforts to disseminate modular models[180], i.e., models researchers can tune to specific research questions with a modest input of data and computation, are promising but still in their infancy.

Despite these advances in analytical throughput and automation, one of the remaining bottlenecks in large-scale biomarker studies is cost. High-throughput and highly multiplexed platforms, while powerful, often come with substantial financial burdens — both in terms of sample processing and data analysis. Computational infrastructure and skilled personnel for large-scale data handling introduce further economic challenges. Thus, while technical barriers are being overcome, economic constraints continue to influence the design and scalability of biomarker studies.

5. Conclusion

The evolution of analytical techniques for multiplex analysis of protein biomarkers has been remarkable, starting from early immunoassay methods and progressing to advanced MS-based techniques and high-throughput affinity-based platforms. These technological advancements have not only increased the multiplexing capacity and sensitivity of protein detection but have also significantly reshaped research methodologies and clinical diagnostics by enabling more comprehensive and precise biomarker discovery. This has led to better understanding and elucidation of disease mechanisms, ultimately facilitating personalized medicine. The future of this field will likely be propelled by groundbreaking advancements in single-cell and spatial proteomics, the integration of multi-omics data, and high-throughput screening technologies. These emerging innovations promise to significantly enhance our ability to decode complex biological systems and uncover novel biomarkers, leading to more accurate disease diagnosis and personalized treatment strategies.

6. Expert Opinion

Advancements in protein biomarker technologies, particularly in MS- and affinity-based platforms, have the potential to revolutionize diagnostics and personalized treatment plans. Improved sensitivity and specificity of these technologies allow for earlier and more accurate disease detection, monitoring, and prognosis. Implementation into clinical practice could lead to better patient outcome and more cost-effective healthcare by providing targeted therapies and reducing unnecessary treatments. However, challenges such as high costs, technical complexity, and the need for large-scale validation trials hinder widespread implementation.

Key areas for improvement include enhancing the reproducibility and scalability of biomarker assays, reducing costs, and increasing the accessibility of advanced technologies. Technical challenges include managing the high dynamic range of proteomes, dealing with sample variability, and addressing the complexity of data analysis. Standardizing protocols, developing more robust and user-friendly analytical tools, and fostering interdisciplinary collaborations can help overcome these limitations.

Further research holds immense potential in uncovering novel biomarkers, understanding disease mechanisms at a deeper level, and developing more sophisticated and individualized therapeutic strategies. While there may never be a definitive endpoint, as scientific inquiry continuously evolves, achieving comprehensive proteomic and multi-omic integration could significantly advance our understanding and management of complex diseases.

While protein biomarker discovery remains a highly promising area, integrating multi-omics approaches (including genomics, metabolomics, and lipidomics) offers even greater potential. Single-cell and spatial proteomics may provide unprecedented insights into cellular heterogeneity and disease microenvironments. These approaches, combined with advanced computational techniques and data integration strategies, could become standard in future research and clinical practice.

In the future, the field is likely to shift towards more high-throughput, automated, and integrative approaches for biomarker discovery. Standard procedures will become more streamlined, incorporating advanced data analysis and machine learning for better predictive modeling and decision-making. While this evolution will bring higher efficiency and precision, it may challenge the simplicity and accessibility of current diagnostic methods, necessitating continuous education and adaptation within clinical and research communities.

Article highlights.

• Protein biomarkers are crucial in clinical diagnostics, disease progression monitoring, and personalized therapeutic guidance, improving patient outcomes in various medical fields.

• Multiplexed protein measurements enable the analysis of multiple biomarkers in a single sample, providing a comprehensive view of biological phenotypes and enhancing diagnostic precision.

• Recent developments in mass spectrometry and affinity-based techniques have significantly improved the accuracy and efficiency of protein biomarker detection.

• Selecting quantitative platforms should align with specific research objectives, recognizing that correlations between platforms may differ and require careful interpretation. The integration of multiple platforms leverages their unique strengths, enhancing the reliability and scope of biomarker development.

• Emerging approaches such as multi-omics integration, single-cell and spatial proteomics, and high-throughput screening promise to unravel complex disease mechanisms and identify novel therapeutic targets.