Olink Proteomics: A Practical Guide to Panel Selection and Data Interpretation

Olink proteomics uses proximity extension assays to measure hundreds of proteins simultaneously in plasma, serum, cerebrospinal fluid, and other biofluids. This guide provides a decision framework for selecting Olink panels based on research goals, sample type, and throughput, followed by a workflow for data preprocessing and interpretation. The content is written for students, researchers, analysts, and life-science professionals who need practical guidance for biomarker discovery studies.

Understanding the Olink Platform and Its Place in Proteomics

Olink technology relies on proximity extension assays, where pairs of antibodies labeled with complementary DNA oligonucleotides bind to their target proteins. When both antibodies bind close together on the same protein, the DNA strands hybridize and can be amplified and quantified using real-time PCR or next-generation sequencing. This approach converts protein detection into a nucleic acid readout, which provides high specificity because both antibodies must bind for a signal to be generated.

The platform has been applied across oncology, cardiovascular disease, respiratory conditions, and immune-related disorders. A review of Olink technology in clinical biomarker discovery describes its utility for identifying biomarkers and potential therapeutic targets, while also noting that each proteomics method has limitations and that combining multiple approaches strengthens validation [5]. The platform has been used to measure 2911 proteins in plasma samples from 22416 UK Biobank participants to predict aortic aneurysm and dissection risk [6], and to profile 92 inflammation-related proteins in pleural effusions for tuberculosis diagnosis [7].

Compared with mass spectrometry-based approaches, Olink offers a different tradeoff. A direct comparison of Olink Explore 3072 with peptide fractionation-based mass spectrometry on 88 plasma samples found complementary proteome coverage and high precision on both platforms, but only moderate quantitative agreement between them, with a median correlation of 0.59 [19]. This means the two technologies measure overlapping but not identical aspects of the proteome, and the choice of platform affects downstream findings.

Panel Selection Decision Framework

Selecting the right Olink panel requires matching the panel content to the biological question, the sample type, and the number of samples to be processed. The decision table below summarizes the main considerations.

Matching Panel Content to Biological Questions

The first decision is whether the research question is hypothesis-driven or discovery-oriented. Hypothesis-driven studies benefit from smaller targeted panels such as the Target 96 Inflammation panel, which quantifies 92 inflammation-related proteins. This panel has been used to identify inflammatory protein signatures associated with vascular cognitive impairment in diabetes [12], to profile plasma from children with autism spectrum disorder [20], and to assess cytokine and immunologic checkpoint molecules in allergen immunotherapy [13].

Discovery-oriented studies benefit from broader panels such as the Explore 3072, which measures thousands of proteins. The Explore 3072 platform was used to identify 33 differentially abundant plasma proteins in patients with amyotrophic lateral sclerosis compared with controls, leading to a diagnostic model with high accuracy [21]. The Explore 384 Inflammation panel was used to identify 225 proteins with significantly altered expression in placental-mediated fetal growth restriction [9].

Sample Type and Volume Considerations

Sample type determines which panels are appropriate and how much material is needed. Plasma and serum are the most common sample types and work well with most panels. Cerebrospinal fluid requires panels with demonstrated performance in that matrix, such as the neurology and inflammation panels used in a study of aneurysmal subarachnoid hemorrhage [11]. Pleural effusion samples have been profiled with the Target 96 Inflammation panel [7].

Sample volume is a practical constraint. Targeted panels typically require less input material than broad Explore panels. Researchers should verify the minimum volume requirements for their chosen panel before collecting samples, especially for precious or limited specimens such as cerebrospinal fluid or pediatric samples.

Throughput and Cost Tradeoffs

Throughput considerations include the number of samples, the number of proteins measured, and the cost per sample. Targeted panels are less expensive per sample and produce smaller data matrices that are easier to analyze. Explore panels provide more data per sample but at higher cost and with greater computational demands.

For studies with large cohorts, such as the UK Biobank analysis of 22416 participants [6], the Explore platform provides the throughput needed for population-scale proteomics. For smaller mechanistic studies, targeted panels offer sufficient coverage with simpler analysis workflows.

Data Preprocessing Workflow

Once samples have been processed on the Olink platform, the data require several preprocessing steps before statistical analysis. The workflow below outlines the essential steps.

Quality Control and Normalization

Olink data are reported in Normalized Protein eXpression units, which are log2-scaled values. The platform includes internal controls for each sample, including incubation controls, extension controls, and detection controls. These controls allow assessment of assay performance and identification of failed samples.

The first quality control step is to review sample-level metrics. Samples that fail internal control thresholds should be flagged and potentially excluded. The second step is to examine the distribution of protein values across samples to identify outliers or batch effects. The third step is to apply normalization if samples were processed in multiple batches or runs.

Handling Missing Values

Missing values in Olink data can arise from proteins below the limit of detection or from technical failures. The proportion of missing values should be assessed for each protein. Proteins with high missingness across samples may need to be excluded from analysis. For proteins with low missingness, imputation methods can be applied, but the choice of imputation method should be documented and justified.

Batch Effect Correction

Studies that process samples in multiple batches require batch effect correction. The Olink platform includes bridge samples or reference samples that can be used to adjust for batch differences. Statistical methods such as ComBat or linear mixed models can also be applied. The choice of method depends on the study design and the number of batches.

Statistical Analysis and Interpretation

Differential Expression Analysis

Differential expression analysis compares protein levels between groups, such as disease versus control. Common approaches include t-tests, Wilcoxon rank-sum tests, or linear regression models with appropriate covariates. Multiple testing correction is essential because hundreds or thousands of proteins are tested simultaneously. The Benjamini-Hochberg procedure for controlling the false discovery rate is commonly used.

In the tuberculous pleural effusion study, differential analysis identified 43 proteins with distinct expression levels between tuberculous and malignant pleural effusions, and 33 between tuberculous and parapneumonic effusions [7]. In the fetal growth restriction study, 225 proteins showed significantly altered expression between affected pregnancies and normal controls [9].

Machine Learning Approaches

Machine learning methods are increasingly used to build diagnostic or predictive models from Olink data. Common approaches include logistic regression, random forests, gradient boosting, and least absolute shrinkage and selection operator modeling. The choice of method depends on the sample size, the number of proteins, and the goal of the analysis.

The aortic aneurysm study used Cox regression and light gradient-boosting machine to identify four key predictive proteins and construct a risk prediction model [6]. The sarcopenia study used Gaussian naive Bayes classifiers for single-omics models and logistic regression for combined models [8]. The breast cancer study used differentially expressed proteins to construct a diagnostic model that was validated in an independent cohort [15].

Model Validation

Model validation is critical for assessing generalizability. Internal validation can be performed using cross-validation or bootstrap resampling. External validation requires testing the model in an independent cohort. The breast cancer study validated its diagnostic model in an independent cohort of 111 patients and 95 healthy controls [15]. The amyotrophic lateral sclerosis study replicated findings in an independent cohort of 48 patients and 75 controls [21].

Pathway and Enrichment Analysis

After identifying differentially expressed proteins, pathway and enrichment analyses can provide biological context. Gene Ontology enrichment analysis and pathway analysis tools can identify biological processes and pathways that are overrepresented among the significant proteins. The fetal growth restriction study found enrichment in pathways related to placental dysfunction, inflammatory responses, and oxidative stress [9]. The vascular cognitive impairment study found enrichment in cytokine-cytokine receptor interaction, TNF signaling, and chemokine signaling pathways [12].

Records and Measurements

Maintaining detailed records is essential for reproducible Olink proteomics studies. The following records should be documented for each study.

Sample Collection Records

Sample collection records should include the date and time of collection, the sample type, the collection tube type, the processing protocol, and the storage conditions. For plasma, the anticoagulant used and the time between collection and processing can affect protein measurements. For cerebrospinal fluid, the collection method and the presence of blood contamination should be recorded.

Assay Run Records

Assay run records should include the panel used, the batch or run identifier, the sample plate layout, and the internal control values. These records allow assessment of batch effects and technical variability.

Data Processing Records

Data processing records should document the software versions, the normalization method, the missing value handling approach, and the batch correction method. These records ensure that the analysis can be reproduced by other researchers.

Common Failure Patterns

Several common failure patterns can compromise Olink proteomics studies.

Inadequate Sample Size

Studies with small sample sizes may lack statistical power to detect true differences, especially after multiple testing correction. The cerebrospinal fluid study of aneurysmal subarachnoid hemorrhage included only six patients and six controls, which limits the generalizability of the findings [11]. Researchers should perform power calculations before starting a study.

Overfitting in Machine Learning Models

Machine learning models with many proteins and few samples are prone to overfitting. The sarcopenia study used a discovery cohort of 80 participants and a validation cohort of 60 participants, which provides some protection against overfitting [8]. Researchers should use cross-validation and external validation to assess model performance.

Ignoring Batch Effects

Failing to account for batch effects can introduce spurious differences between groups, especially if cases and controls are processed in different batches. Study designs should randomize samples across batches whenever possible.

Overinterpretation of Exploratory Findings

Exploratory studies generate hypotheses that require validation. The allergen immunotherapy study found no significant correlations between pre-treatment cytokine levels and treatment outcome, highlighting the importance of replication and the risk of false positives in exploratory analyses [13].

Limitations and Interpretation Caveats

Olink proteomics has several limitations that affect data interpretation.

Relative Quantification

Olink data provide relative quantification instead of absolute protein concentrations. The Normalized Protein eXpression values are log2-scaled and reflect relative abundance across samples. Comparisons across studies require careful attention to normalization methods.

Moderate Agreement with Mass Spectrometry

The moderate quantitative agreement between Olink and mass spectrometry platforms means that findings from one platform may not be directly reproducible on the other [19]. Researchers should consider whether their findings are platform-specific or reflect true biological differences.

Protein Panel Content

The proteins included in each panel are preselected. Proteins not included in the panel cannot be measured, which limits the scope of discovery. The Target 96 Inflammation panel measures 92 proteins, which represents a small fraction of the human proteome.

Validation Requirements

Biomarker candidates identified through Olink proteomics require validation using independent methods such as enzyme-linked immunosorbent assay. The tuberculous pleural effusion study validated four key proteins using ELISA [7]. The psoriasis study also validated key differentially expressed proteins using ELISA [17].

Quality Controls and Reproducibility

Quality controls are essential for producing reliable Olink data.

Internal Controls

The Olink platform includes multiple internal controls for each sample. Incubation controls assess the efficiency of the proximity extension reaction. Extension controls assess the efficiency of the extension step. Detection controls assess the efficiency of the amplification and detection steps. Samples with abnormal control values should be flagged.

Replicates

Technical replicates can assess assay precision. Biological replicates are necessary to assess biological variability. The cerebrospinal fluid study reported good intra- and inter-group reproducibility based on principal component analysis [11].

Data Sharing and Reproducibility

Following the FAIR Guiding Principles for scientific data management ensures that data are findable, accessible, interoperable, and reusable [4]. Researchers should deposit raw and processed data in public repositories such as those maintained by the National Center for Biotechnology Information [2]. Data sharing policies, such as the NIH Genomic Data Sharing Policy, may apply to certain types of research [3]. Training resources for data management and analysis are available through the European Bioinformatics Institute [1].

Safety and Regulatory Context

Olink proteomics research involving human subjects must comply with applicable regulations and ethical standards. Researchers should obtain appropriate ethical approval before collecting samples. Data sharing must protect participant privacy and comply with applicable data protection regulations.

The NIH Genomic Data Sharing Policy provides guidance on data sharing for NIH-funded research [3]. Researchers should review their funding agency requirements and institutional policies before starting a study.

Professional Escalation Criteria

Researchers should seek expert assistance in the following situations.

Statistical Consultation

Consult a biostatistician when designing the study, performing power calculations, or analyzing complex data structures. Statistical expertise is especially important for machine learning analyses and for studies with longitudinal or repeated measures designs.

Bioinformatics Support

Consult a bioinformatician for data preprocessing, batch effect correction, and pathway analysis. These steps require specialized software and computational expertise.

Clinical Collaboration

Collaborate with clinicians when interpreting findings in a disease context. Clinical expertise is necessary to assess the biological plausibility of findings and to plan validation studies.

Building a Cross-Platform Validation Workflow for Olink Findings

A common weakness in Olink proteomics studies is the failure to plan for validation before the discovery phase begins. Researchers often complete a full Olink experiment, identify candidate proteins, and only then consider how to confirm the findings. This approach creates avoidable delays and can produce candidates that cannot be validated because the appropriate samples, assays, or statistical plans were not prepared in advance. A structured cross-platform validation workflow addresses this problem by embedding validation planning into the earliest stages of study design.

Step 1: Define the Validation Hierarchy Before Sample Collection

The first decision is which validation method will serve as the primary confirmation for each candidate protein. Enzyme-linked immunosorbent assay is the most widely used orthogonal method in Olink studies. The tuberculous pleural effusion study used ELISA to confirm four key inflammatory proteins identified by Olink, including IFN-gamma, CXCL9, TNF-beta, and PD-L1, with a combined area under the curve of 0.963 [7]. The psoriasis study similarly validated key differentially expressed proteins using ELISA after Olink profiling [17].

Mass spectrometry represents a second validation tier. A direct comparison of Olink Explore 3072 with peptide fractionation-based mass spectrometry on 88 plasma samples found complementary proteome coverage and high precision on both platforms, but only moderate quantitative agreement, with a median correlation of 0.59 [19]. This moderate agreement means that mass spectrometry validation is informative but should not be expected to reproduce Olink values exactly. Researchers should interpret mass spectrometry confirmation as evidence of biological presence instead of quantitative equivalence.

The validation hierarchy should be documented in the study protocol before any samples are processed. This documentation should specify which proteins will be validated, which method will be used for each protein, and what threshold will define successful validation.

Step 2: Reserve Samples for Validation During the Discovery Phase

Sample scarcity is the most common reason validation fails. When a study uses all available material for Olink profiling, there is nothing left for confirmatory assays. The solution is to split samples at the time of collection and store a dedicated validation aliquot.

For plasma and serum studies, collect an additional aliquot at the same venipuncture and store it under identical conditions. For cerebrospinal fluid studies, where volumes are limited, the validation aliquot may need to be smaller, and the validation plan should account for this constraint. The aneurysmal subarachnoid hemorrhage study profiled cerebrospinal fluid from six patients and six controls using both neurology and inflammation panels [11], demonstrating that multiplexed profiling from limited material is feasible, but this leaves minimal volume for orthogonal validation.

The validation aliquot should be frozen without thawing until the validation assay is performed. Repeated freeze-thaw cycles degrade proteins and can produce false negative validation results. Each aliquot should be labeled with the same study identifier as the discovery sample, and the storage location and temperature should be recorded in the sample collection log.

Step 3: Select Validation Assays Based on Protein Characteristics

Not all proteins identified by Olink are equally suitable for validation by a given method. The choice of validation assay should consider the protein's expected abundance, the sample matrix, and the availability of commercial assays.

For high-abundance proteins in plasma or serum, ELISA kits are generally available and perform reliably. For low-abundance proteins, more sensitive assays may be required, or the validation may need to be performed in a matrix where the protein is more concentrated. The breast cancer study identified INPP1 and ARHGAP25 as serum biomarkers and validated the diagnostic model in an independent cohort of 111 patients and 95 healthy controls [15], demonstrating that validation can succeed when the assay is matched to the protein and matrix.

For proteins without commercial ELISA kits, alternative validation approaches include targeted mass spectrometry assays or multiplexed bead-based immunoassays. The choice should be documented in the validation plan, along with the expected limit of detection and the assay's demonstrated performance in the relevant sample type.

Step 4: Predefine Validation Success Criteria

Validation success should be defined before the validation assay is performed. This prevents post hoc interpretation that can bias the assessment. The success criteria should address both statistical significance and direction of effect.

A common approach is to require that the validation assay reproduce the direction of effect observed in the Olink data and reach statistical significance at a predefined threshold. The sarcopenia study used a validation cohort of 30 sarcopenic and 30 non-sarcopenic participants to evaluate models built in a discovery cohort of 80 participants [8], demonstrating the importance of independent validation cohorts with adequate sample size.

The success criteria should also specify how discordant results will be handled. A protein that shows a significant effect in Olink data but no effect in the validation assay may indicate a platform-specific artifact, a false positive in the discovery analysis, or a validation assay that lacks sufficient sensitivity. The protocol should specify whether discordant proteins will be excluded, re-tested with an alternative method, or reported with a caveat.

Step 5: Document Platform Agreement Metrics

When both Olink and a validation method are used on the same samples, the agreement between platforms should be quantified and reported. The comparison study between Olink Explore 3072 and mass spectrometry reported a median correlation of 0.59 between platforms, with an interquartile range of 0.33 to 0.75 [19]. This wide range indicates that agreement varies substantially across proteins.

For each validated protein, record the Olink value, the validation assay value, and the correlation or concordance measure. This documentation allows researchers to identify proteins where platform agreement is poor and to interpret discordant results in context. The comparison study also introduced a publicly available tool for peptide-level analysis of platform agreement [19], which can help clarify cross-platform discrepancies.

Step 6: Integrate Validation Results into the Final Model

Validation results should inform the final biomarker panel, also confirm individual proteins. The breast cancer study combined INPP1 and ARHGAP25 into a diagnostic model that achieved an area under the curve of 0.8458 in the discovery cohort and 0.8506 in the validation cohort [15]. The model retained efficacy in early-stage detection with an area under the curve of 0.7598 [15].

When integrating validation results, consider whether the validated proteins improve model performance when added to clinical variables. The aortic aneurysm study integrated four key predictive proteins with demographic factors and achieved an area under the curve of 0.777, compared with 0.740 for the demographic model alone [6]. This improvement demonstrates the added value of validated proteomic biomarkers over traditional risk factors.

Common Validation Failure Patterns

Several recurring problems undermine validation efforts. The first is attempting validation with too few samples. Validation cohorts should be sized to detect the effect size observed in the discovery phase with adequate statistical power. The sarcopenia study used a validation cohort of 60 participants [8], while the breast cancer study used 206 participants [15]. Smaller validation cohorts risk false negative results.

The second failure pattern is using the same samples for discovery and validation. This inflates performance estimates and does not demonstrate generalizability. Validation must use independent samples, preferably collected at a different time or site.

The third failure pattern is ignoring the moderate agreement between platforms. When Olink and mass spectrometry show a median correlation of 0.59 [19], researchers should expect some proteins to show discordant results and should not interpret every discordance as a validation failure.

The fourth failure pattern is failing to document validation methods in sufficient detail for reproduction. The validation protocol should specify the assay manufacturer, lot number, dilution factor, incubation conditions, and analysis method. This documentation supports the FAIR Guiding Principles for scientific data management, which require that data and methods be findable, accessible, interoperable, and reusable [4].

Records for Validation Studies

Maintain a validation log that records the following for each candidate protein: the Olink panel and protein identifier, the validation method, the assay manufacturer and catalog number, the sample identifiers used for validation, the raw validation values, the statistical test and result, and the final validation decision. This log should be stored with the study data and referenced in any publications or reports.

The validation log should also record any samples that were excluded from validation and the reason for exclusion. Sample exclusions due to insufficient volume, hemolysis, or failed assay controls should be documented to allow assessment of potential bias.

Professional Escalation for Validation Challenges

Consult a clinical chemist or assay specialist when selecting validation assays for proteins with known matrix effects or when commercial assays are unavailable. Consult a biostatistician when designing the validation analysis, particularly for studies with small validation cohorts or when multiple testing correction is required. If validation consistently fails across multiple proteins, consult a proteomics methodologist to assess whether the Olink data or the validation approach requires revision.

When validation results will be used to support regulatory submissions or clinical decisions, consult the relevant regulatory guidance and institutional review processes before beginning the validation work. The NIH Genomic Data Sharing Policy provides guidance on data sharing for NIH-funded research [3], and researchers should review their funding agency requirements before initiating validation studies.

Frequently Asked Questions

What is the difference between Olink Target and Olink Explore panels?

Target panels measure a fixed set of approximately 92 proteins using real-time PCR readout. Explore panels measure hundreds to thousands of proteins using next-generation sequencing readout. Explore panels provide broader coverage but require more complex data analysis and higher cost per sample.

How much sample volume is needed for Olink analysis?

Sample volume requirements vary by panel. Targeted panels typically require less input material than Explore panels. Researchers should verify the minimum volume requirements for their chosen panel before collecting samples, especially for limited specimens such as cerebrospinal fluid.

Can Olink data be compared with mass spectrometry data?

Quantitative agreement between Olink and mass spectrometry platforms is moderate, with a median correlation of 0.59 reported in one comparison study [19]. The platforms provide complementary coverage, and combining both approaches can provide more comprehensive proteome profiling.

How should missing values in Olink data be handled?

Assess the proportion of missing values for each protein. Exclude proteins with high missingness across samples. For proteins with low missingness, apply imputation methods and document the choice. The imputation method should be appropriate for the data structure and the downstream analysis.

What is the Normalized Protein eXpression unit?

Normalized Protein eXpression is a log2-scaled unit used by Olink to report relative protein abundance. The values are normalized using internal controls and reference samples to allow comparisons across samples and runs.

How many proteins can be measured with the Olink Explore 3072 panel?

The Olink Explore 3072 panel measures approximately 3072 proteins. In one comparison study, 1129 proteins were analyzed with both Olink Explore 3072 and mass spectrometry [19].

What validation methods are recommended for Olink biomarker candidates?

Biomarker candidates should be validated using independent methods such as enzyme-linked immunosorbent assay. Several studies have used ELISA to validate Olink findings, including studies of tuberculous pleural effusion [7] and psoriasis [17].

Can Olink proteomics be used for longitudinal studies?

Yes, Olink proteomics can be used for longitudinal studies. The psoriasis study quantified 92 inflammation-related proteins in plasma samples collected before and during secukinumab therapy [17]. Longitudinal designs require careful attention to batch effects and sample handling across time points.

Related Bioinformatics Guides

• Data Sharing and Privacy in Genomic Research
• Alternative Splicing Analysis from RNA-Seq Data
• Genomic Selection in Animal Breeding
• Gene Ontology (GO) and Enrichment Analysis
• The KEGG Database and Pathway Analysis

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Methods and clinical biomarker discovery for targeted proteomics using Olink technology.. Proteomics. Clinical applications, 2024.
• Plasma proteomics profiles predict the risk of future aortic aneurysm and aortic dissection.. International journal of surgery (London, England), 2025.
• Identification of a Four-Biomarker Panel for the Diagnosis of Tuberculous Pleural Effusion Using Olink Proteomics.. Journal of inflammation research, 2026.
• Integrated Proteomic and Metabolomic Profiling for Developing Novel Plasma-Based Diagnostic Models of Sarcopenia.. Journal of cachexia, sarcopenia and muscle, 2026.
• Investigation of potential protein biomarkers for the screening of placental-mediated fetal growth restriction disorders using targeted proteomics Olink technology.. Frontiers in immunology, 2025.
• A multi-layer functional genomic analysis to understand noncoding genetic variation in lipids.. American journal of human genetics, 2022.
• Multi-targeted olink proteomics analyses of cerebrospinal fluid from patients with aneurysmal subarachnoid hemorrhage.. Proteome science, 2024.
• OLINK proteomics identifies inflammatory protein signatures associated with vascular cognitive impairment in diabetes.. Frontiers in immunology, 2026.
• Cytokines and immunologic checkpoint molecules in predicting success of allergen immunotherapy.. 2026.
• Method optimization of protein and nucleic acid extraction from skin tape strip samples for biomarker analysis
• Clinical validation of PEA-driven Olink proteomic discovery: INPP1 and ARHGAP25 serum biomarkers improve early breast cancer diagnosis.. 2026.
• Olink Proteomics in Heart Failure: A Comprehensive Review of Applications From Biomarker Discovery to Pathophysiological Insights and Clinical Implementation.. 2026.
• Longitudinal plasma proteomics identifies diagnostic and response-associated inflammatory and immune biomarkers in psoriasis following secukinumab therapy.. 2026.
• Plasma Proteomics Identifies Insulin Resistance Signatures Predictive of Cardiometabolic Disease Risk. 2026.
• Comparative evaluation of Olink Explore 3072 and mass spectrometry with peptide fractionation for plasma proteomics. Communications Chemistry, 2025.
• Novel Inflammatory Biomarkers for Autism Spectrum Disorder Detected by Plasma Olink Proteomics. Children, 2025.
• A plasma proteomics-based candidate biomarker panel predictive of amyotrophic lateral sclerosis. Nature Medicine, 2025.
• Combined Plasma Olink Proteomics and Transcriptomics Identifies CXCL1 and TNFRSF12A as Potential Predictive and Diagnostic Inflammatory Markers for Acute Kidney Injury. Inflammation, 2024.
• Serum Olink Targeted Proteomics Identifies IL-17A as a Prospective Inflammatory Marker for the Prediction and Diagnosis of Kawasaki Disease. Journal of Inflammation Research, 2025.
• Predicted plasma proteomics from genetic scores and treatment outcomes in major depression: a meta-analysis. European Neuropsychopharmacology, 2025.
• Bone marrow-based highly sensitive proteomics profiling reveals valuable biomarkers for pediatric B-cell acute lymphoblastic leukemia. Cancer Letters, 2026.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.