Pathway Enrichment Analysis for Proteomics: Tools and Interpretation

Pathway enrichment analysis is a computational approach that determines whether proteins from your proteomics experiment are overrepresented in known biological pathways, such as KEGG metabolic pathways, gene ontology terms, or curated signaling cascades. This article provides a practical framework for selecting tools, running enrichment analyses on proteomics datasets, and interpreting results with appropriate caution. The workflow covers data preparation, tool selection, statistical controls, reproducibility practices, and common interpretation errors that lead to false biological conclusions.

At a Glance

Pathway enrichment analysis answers a specific question: given a list of differentially abundant proteins or a ranked protein list, which biological pathways contain more proteins from your dataset than expected by chance? The table below summarizes the main analysis approaches, their data requirements, and typical use cases in proteomics.

The choice between over-representation analysis and rank-based enrichment is the first major decision. Over-representation analysis requires a binary call of significant versus non-significant proteins, which discards quantitative information. Rank-based methods preserve the full distribution of protein measurements and can detect pathway-level shifts that are invisible when proteins are filtered by individual significance thresholds.

Proteomics Data Inputs and Preprocessing

Protein Quantification Tables

Pathway enrichment begins with a protein quantification table, typically generated from mass spectrometry-based proteomics workflows. Data-independent acquisition methods produce quantitative readouts of protein abundance across samples, while data-dependent acquisition approaches provide deeper identification at the cost of more missing values. The choice of acquisition method affects downstream enrichment analysis because missingness patterns differ substantially between approaches.

Before running enrichment, you must decide which proteins enter the analysis. For over-representation analysis, you need a defined list of differentially abundant proteins and a background list of all proteins detected in the experiment. The background list is critical. Using only the significant proteins without a proper background inflates enrichment scores because the reference set is too small. The background should be all proteins quantified in your experiment, not the entire human proteome, unless you have a specific reason to use the full proteome.

Handling Missing Values

Proteomics datasets frequently contain missing values from proteins below detection limits or from stochastic sampling in data-dependent acquisition. Missingness affects enrichment results because proteins with many missing values may be excluded from statistical tests, biasing the protein list that enters enrichment analysis. Machine learning approaches for proteomics emphasize careful missing-data handling before any downstream interpretation, since the choice of imputation method can change which proteins are called significant.

For enrichment analysis, record the number of proteins quantified in each sample and the proportion of missing values per protein. If a pathway contains many proteins with high missingness, that pathway may appear depleted or enriched depending on how missing values are handled. Document the missingness threshold used to retain proteins and report this in methods sections.

Batch Effects and Normalization

Batch effects from sample processing dates, instrument runs, or reagent lots can introduce systematic variation that mimics biological pathway signals. Normalization methods such as median centering or variance stabilization reduce technical variation, but they do not eliminate all batch structure. If samples were processed in multiple batches, include batch as a covariate in differential abundance testing before enrichment. The reproducibility of functional enrichment results depends heavily on these statistical choices, and different normalization strategies can lead to different enriched pathways from the same raw data.

Core Principles of Enrichment Statistics

Over-Representation Analysis

Over-representation analysis applies a hypergeometric test or Fisher exact test to determine whether a pathway contains more proteins from your significant list than expected by chance. The test requires four numbers: proteins in your list that belong to the pathway, proteins in your list outside the pathway, background proteins in the pathway, and background proteins outside the pathway.

The main limitation of over-representation analysis is the binary threshold. Proteins with a p-value of 0.049 and 0.051 are treated identically in one case and completely differently in another, even though the biological evidence is nearly the same. This threshold dependence makes over-representation analysis sensitive to the statistical power of the differential abundance test, which varies with sample size, effect size, and missing data patterns.

Rank-Based Enrichment Analysis

Gene set enrichment analysis and related rank-based methods avoid binary thresholds by ranking all measured proteins by a test statistic, such as the t-statistic from a differential abundance test or a fold change. The method then walks down the ranked list and calculates an enrichment score that reflects whether proteins from a given pathway cluster at the top or bottom of the ranking.

Rank-based methods detect coordinated shifts in pathway activity even when individual proteins do not reach significance. This is valuable in proteomics because protein-level changes are often modest in magnitude, especially for regulatory proteins that change abundance by less than twofold. The TeaProt tool integrates gene set enrichment analysis with upstream transcription factor analysis and downstream pathway interpretation, allowing users to run rank-based enrichment against multiple gene set resources including MSigDB and curated post-translational modification databases.

Network-Based Approaches

Standard enrichment methods treat pathways as flat gene sets and disregard interactions among proteins within a pathway. This simplification can produce partial or inaccurate inference because it ignores the topology of signaling cascades and metabolic networks. Network-based enrichment methods incorporate interaction structure to identify pathway disturbances driven by topologically important proteins.

Causal Disturbance Analysis uses a graph centrality model called Source/Sink to quantify whether a protein acts primarily as a receiver or sender of biological information within a pathway. This approach prioritizes proteins that are likely to disturb pathway functionality and combines topological evidence with classical over-representation analysis. In evaluations with real and synthetic data, this method identified pathway enrichments not observable through standard enrichment models, particularly for pathways where central hub proteins changed expression while peripheral proteins remained stable.

Tool Selection for Proteomics Enrichment

General-Purpose Enrichment Tools

ClusterProfiler is an R package that supports over-representation analysis and gene set enrichment analysis across multiple databases including KEGG, Gene Ontology, and Reactome. It accepts protein identifiers and provides visualization functions for enrichment results. The package is widely used in proteomics publications and integrates with standard Bioconductor workflows.

Enrichr is a web-based tool that runs enrichment against dozens of gene set libraries simultaneously. It is useful for exploratory analysis but has limitations for proteomics because many of its gene set libraries are derived from transcriptomics data. Protein identifiers must be mapped to gene symbols before submission.

DAVID provides functional annotation and enrichment analysis with a focus on gene ontology terms and KEGG pathways. It accepts gene lists and performs over-representation analysis with a background option. DAVID is best suited for hypothesis generation instead of final pathway interpretation.

Proteomics-Specific Tools

TeaProt is an online Shiny tool designed for proteomics analysis that integrates upstream transcription factor enrichment with downstream pathway analysis. It annotates uploaded data with drug-gene interactions, subcellular localizations, phenotypic functions, gene-disease associations, and enzyme-gene interactions. TeaProt enables gene set enrichment analysis against resources including MSigDB, CHEA, and the underrepresented post-translational modification database urPTMdb.

The urPTMdb database provides curated lists of proteins reported as substrates of underrepresented post-translational modifications. Enrichment of these protein lists in proteomics datasets can reveal unexpected post-translational modification regulation that standard pathway databases miss. In a demonstration using a Western diet-induced tongue proteome dataset, TeaProt and urPTMdb revealed altered cellular processes associated with energy metabolism, interferon responses, adipogenesis, and transcription regulation through PPARG and CEBPA.

pathlinkR is an R package available from Bioconductor that streamlines pathway enrichment and network-based analysis for human omics data. It produces publication-quality figures and integrates enrichment results with network visualization. The package is designed for transcriptomic data but accepts protein-level inputs after identifier mapping.

InCroMAP supports integrated enrichment analysis and pathway-centered visualization across metabolomics, proteomics, transcriptomics, and genomics data. This tool is valuable when you have matched multi-omics data from the same samples and want to identify pathways that show coordinated changes across molecular layers.

Database Choices

KEGG pathways are the most commonly used database in proteomics enrichment analysis. KEGG provides curated metabolic and signaling pathways with cross-references to enzyme and transporter annotations. The database is updated regularly, and pathway definitions change between releases, so record the database version used in your analysis.

Gene Ontology provides three structured vocabularies covering biological processes, molecular functions, and cellular components. GO enrichment is useful for identifying broad functional categories but produces many redundant terms because of the hierarchical structure. Tools that collapse redundant GO terms, such as ClusterProfiler with simplify functions, reduce the interpretation burden.

Reactome provides curated human pathways with detailed reaction steps. It is particularly useful for signaling pathways and has better coverage of non-metabolic processes than KEGG. Reactome also provides pathway hierarchy information that helps interpret enrichment results in context.

Step-by-Step Protocol for Running Enrichment on Proteomics Data

Step 1: Define the Analysis Question

Write down the biological question before running any software. Are you asking which pathways are activated in disease compared to control? Are you looking for pathway differences between treatment groups? Are you testing a specific hypothesis about a metabolic pathway? The analysis question determines whether you need over-representation analysis on a significant protein list or rank-based enrichment on all measured proteins.

Step 2: Prepare the Protein List and Background

Export the differential abundance results from your proteomics analysis. For over-representation analysis, create two lists: the significant proteins based on your chosen thresholds and the background list of all proteins quantified in the experiment. For rank-based enrichment, create a ranked table with protein identifiers and a test statistic such as the t-statistic or signed log p-value.

Map protein identifiers to a consistent format. Most enrichment tools accept gene symbols or Entrez IDs. UniProt accession numbers must be converted before analysis. Record the mapping rate because proteins that fail to map are excluded from enrichment and can bias results if they cluster in specific pathways.

Step 3: Select the Enrichment Method

Choose over-representation analysis when you have a clear significant protein list and a well-defined background. Choose rank-based enrichment when you want to detect coordinated pathway shifts without threshold dependence. Consider network-based methods when you have prior knowledge that pathway topology matters for your biological question.

Step 4: Run Enrichment Against Multiple Databases

Run enrichment against at least two databases, typically KEGG and Gene Ontology biological process. Cross-database consistency strengthens confidence in pathway findings. A pathway enriched in both KEGG and GO with overlapping member proteins is more likely to reflect a real biological signal than a pathway enriched in only one database.

Step 5: Apply Multiple Testing Correction

Enrichment analysis tests hundreds or thousands of pathways simultaneously. Apply multiple testing correction such as Benjamini-Hochberg false discovery rate control. Report adjusted p-values instead of raw p-values. The choice of correction method affects which pathways pass significance thresholds, and different statistical choices can change the final enrichment results.

Step 6: Visualize and Interpret Results

Generate dot plots, bar charts, or pathway diagrams to visualize enrichment results. Examine the direction of change for proteins within enriched pathways. A pathway enriched with upregulated proteins has a different biological meaning than a pathway enriched with downregulated proteins. Rank-based methods provide direction information through the enrichment score sign.

Step 7: Validate with Independent Methods

Confirm key pathway findings using an orthogonal approach. For example, if lipid metabolism pathways are enriched, check whether lipid-related proteins show coordinated changes across the dataset. If a specific signaling pathway is enriched, examine whether upstream regulators and downstream targets both change in the expected direction. The molecular ICU framework emphasizes pathway-focused biomarkers as clinically translatable signatures that preserve biological mechanisms while enabling practical measurement.

Records and Measurements

Documentation Requirements

Record every parameter that affects enrichment results. This includes the database versions, the identifier mapping approach, the background protein list, the significance thresholds for differential abundance, the multiple testing correction method, and the software versions. The FAIR Guiding Principles emphasize that data should be Findable, Accessible, Interoperable, and Reusable, and this applies to analysis parameters as much as raw data.

Maintain a project log that records the date of each analysis run, the input file versions, and the output file locations. This log enables reproducibility when you or collaborators need to revisit the analysis. The reproducibility of functional enrichment depends on statistical choices, and without complete documentation, results cannot be independently verified.

Quality Control Metrics

Track the number of proteins identified in each sample, the number of proteins passing quality filters, and the number of proteins entering enrichment analysis. A large drop between identification and enrichment input indicates that filtering criteria are too stringent or that identifier mapping failed for many proteins.

Record the proportion of proteins that map to pathway databases. If more than 20 percent of proteins fail to map, the enrichment results may be biased because unmapped proteins are excluded from pathway membership calculations. Check whether unmapped proteins are enriched for specific functional categories, which would indicate systematic bias.

Analysis Output Records

Save the complete enrichment output tables, beyond the significant pathways. These tables include the number of proteins from your dataset in each pathway, the expected number by chance, the enrichment ratio, and the adjusted p-value. Store the ranked protein lists used for rank-based enrichment because these are required for re-analysis with different parameters.

Common Failure Patterns

Using an Incorrect Background

The most common error in over-representation analysis is using the entire proteome as background instead of the proteins actually measured in the experiment. This inflates enrichment significance because the background is larger than the detectable protein space. A pathway may appear enriched simply because its proteins are more detectable by mass spectrometry, not because they change in abundance.

Ignoring Protein Detectability Bias

Mass spectrometry detects abundant proteins more reliably than low-abundance proteins. Pathways composed of abundant proteins, such as metabolic enzymes and structural proteins, are more likely to show enrichment regardless of biological relevance. Compare enrichment results against a background of detected proteins to partially correct for this bias, but recognize that detectability bias cannot be fully eliminated.

Threshold Dependence in Over-Representation Analysis

Over-representation analysis results change dramatically when the significance threshold for differential abundance changes. A pathway may be enriched at a p-value threshold of 0.05 but not at 0.01, simply because fewer proteins enter the significant list. Test sensitivity by running enrichment at multiple thresholds and examining whether core pathway findings persist.

Treating All Pathways as Independent

Pathway databases contain overlapping and hierarchically related pathways. The same protein can appear in multiple pathways, and pathways share member proteins extensively. This non-independence means that enrichment p-values are correlated, and multiple testing correction that assumes independence is conservative or anti-conservative depending on the overlap structure.

Overinterpreting Single Pathway Hits

A single enriched pathway from one database with a marginal adjusted p-value is weak evidence for biological mechanism. Strong pathway findings should replicate across databases, show coordinated protein changes in the expected direction, and align with the broader biological context of the experiment. The proteomic study of adolescent idiopathic scoliosis demonstrated this principle by showing that enrichment analysis identified muscle weakness and hormone disorder pathways in increased proteins while decreased proteins mapped to immune response pathways, with the overall pattern supporting a coherent biological model.

Interpretation Limits and Biological Context

Enrichment Does Not Prove Causality

Pathway enrichment identifies statistical overrepresentation, not causal mechanism. An enriched pathway may reflect downstream consequences of the true biological driver instead of the driver itself. For example, enrichment of immune response pathways in a disease proteome could indicate either immune activation as a cause or immune infiltration as a consequence of tissue damage.

Pathway Databases Are Incomplete

KEGG and Gene Ontology provide curated knowledge, but they do not cover all biological processes. Novel pathways, tissue-specific mechanisms, and recently discovered interactions may be absent from databases. Absence of enrichment does not mean absence of biological relevance. The underrepresented post-translational modification database urPTMdb was developed specifically because standard databases lacked coverage of certain protein modifications, and its use revealed unexpected regulation that standard analysis missed.

Proteomics Measures Abundance, Not Activity

Protein abundance does not directly measure pathway activity. Post-translational modifications, allosteric regulation, subcellular localization, and protein-protein interactions all modulate pathway function without changing protein abundance. A pathway can be highly active with unchanged protein levels, or inactive despite high protein abundance. Phosphoproteomics and other modification-specific approaches provide complementary information but require separate analysis workflows.

Cross-Study Comparison Challenges

Comparing enrichment results across studies is complicated by differences in experimental design, quantification methods, statistical thresholds, and database versions. The meta-analysis of pathway enrichment across independent and dependent omics datasets requires specialized statistical methods that account for these differences. Without such methods, apparent discrepancies between studies may reflect methodological variation instead of biological differences.

Multi-Omics Integration for Pathway Validation

Combining Proteomics with Transcriptomics

Matched transcriptomics and proteomics data from the same samples provide complementary evidence for pathway activity. Messenger RNA abundance reflects transcriptional regulation while protein abundance reflects the integrated outcome of transcription, translation, and degradation. Pathways showing coordinated changes at both levels are more likely to reflect genuine biological regulation than pathways changing at only one level.

The proteomics-informed transcriptomics approach used in the study of severe fever with thrombocytopenia syndrome virus infection in tick cells generated de novo transcriptomes and proteomes of infected and mock-infected cells. This approach identified key host responses and regulatory pathways by integrating both molecular layers, and interactome analysis of the viral nucleoprotein connected host responses with viral replication.

Integrating Metabolomics Data

Metabolomics provides direct measurements of metabolic pathway products and intermediates. When proteomics enrichment identifies a metabolic pathway, metabolomics data can confirm whether the pathway is actually producing or consuming metabolites as expected. The kefir grain interaction study integrated metabolomics with pathway enrichment analysis using KEGG and revealed that co-culture activated core pathways including ABC transporters, amino acid biosynthesis, and protein digestion and absorption, with dynamic metabolite changes providing functional evidence for the pathway findings.

Network Integration for Mechanistic Insight

Protein-protein interaction networks connect enrichment results to mechanistic hypotheses. The Mendelian randomization study of abdominal aortic aneurysm used protein-protein interaction network analysis alongside pathway enrichment to identify cell types with enriched expression and prioritize therapeutic targets. The sarcopenia study similarly combined protein-protein interaction analysis with pathway enrichment and drug target annotation to identify pathogenic proteins and druggable targets.

Quality Controls and Reproducibility

Statistical Reproducibility

The choice of statistical methods substantially affects enrichment results. A study examining the impact of statistical choices on proteomics functional enrichment found that different normalization methods, differential abundance tests, and enrichment algorithms produced divergent pathway lists from the same underlying data. To assess robustness, run the analysis with at least two different statistical approaches and compare the resulting enriched pathways.

Parameter Sensitivity Analysis

Test how enrichment results change with variations in key parameters. Vary the significance threshold for differential abundance, the minimum protein count per pathway, and the multiple testing correction method. Pathways that remain enriched across parameter variations are more robust than pathways that appear only under specific settings.

Independent Replication

The strongest validation for enrichment findings is independent replication in a separate cohort or experiment. The plasma proteomic study of tuberculosis identified significantly differentially expressed proteins in individuals affected by tuberculosis and showed that proteomic alterations preceded clinical onset and persisted years after diagnosis. Functional annotation indicated involvement of immune and signaling pathways associated with chronic inflammatory responses and host defense mechanisms. These findings gain credibility from the prospective cohort design and the persistence of signals across time points.

Reporting Standards

Report enrichment analysis methods with sufficient detail for replication. Include the software and version, the database and version, the identifier mapping approach, the background definition, the statistical test, the multiple testing correction, and the significance thresholds. The FAIR Guiding Principles provide a framework for making data and analysis reusable, and adopting these practices during data collection enables reproducible and interpretable modeling.

Professional Escalation Criteria

When to Seek Specialized Support

Consult a bioinformatics specialist or biostatistician when your enrichment analysis involves complex experimental designs, such as longitudinal sampling, multiple treatment factors, or paired samples. These designs require statistical models that account for correlation structure, and standard enrichment tools may not handle them appropriately.

When to Question Results

Escalate for expert review when enrichment results conflict with established biological knowledge, when the same analysis produces very different results across software tools, or when pathway findings cannot be reproduced with different parameter settings. Conflicting results may indicate data quality problems, inappropriate statistical methods, or database annotation errors.

When to Reanalyze Data

Reanalyze the data when new versions of pathway databases become available, when you discover errors in the original analysis parameters, or when you obtain additional samples that change the statistical power. The proteome-wide Mendelian randomization study of abdominal aortic aneurysm used multiple independent datasets and sequential analyses including colocalization and summary-data-based Mendelian randomization to verify causal roles of candidate proteins, demonstrating the value of multi-dataset validation.

When to Involve Domain Experts

Pathway interpretation benefits from domain expertise. A biologist familiar with the tissue or disease under study can assess whether enriched pathways make biological sense and whether the direction of protein changes is consistent with known mechanisms. The tonsillar B cell study comparing PFAPA and OSAS identified distinct pathway enrichments including alterations in PTEN gene transcription regulation, circadian gene expression, inflammasome pathways, and angiotensinogen metabolism, and domain expertise was essential for interpreting these findings in the context of pediatric tonsillar pathology.

Safety and Regulatory Context

Data Sharing and Privacy

Proteomics data from human subjects may be subject to data sharing policies and privacy regulations. The NIH Genomic Data Sharing Policy provides a framework for sharing genomic data while protecting participant privacy, and similar principles apply to proteomics data derived from human samples. Before sharing proteomics data or enrichment results, verify that the data collection protocol and consent forms permit the intended sharing.

Clinical Translation Considerations

Pathway enrichment results from research studies should not be used for clinical decision-making without appropriate validation. The molecular ICU framework emphasizes that pathway-focused biomarkers must be validated as clinically translatable signatures before use in patient care. Enrichment findings that suggest drug targets or biomarkers require prospective validation in appropriate clinical cohorts.

Reproducibility Requirements

Funding agencies and journals increasingly require data and code availability for published analyses. The FAIR Guiding Principles provide a framework for making data Findable, Accessible, Interoperable, and Reusable. Adopting these practices during data collection enables reproducible and interpretable modeling and supports the reuse of proteomics datasets for pathway analysis by other researchers.

Frequently Asked Questions

What is the difference between over-representation analysis and gene set enrichment analysis?

Over-representation analysis tests whether a predefined list of significant proteins contains more proteins from a pathway than expected by chance. It requires a binary significant or non-significant call and a background protein list. Gene set enrichment analysis ranks all measured proteins by a test statistic and detects whether pathway proteins cluster at the top or bottom of the ranking. Rank-based methods preserve quantitative information and can detect coordinated pathway shifts even when individual proteins do not reach significance thresholds.

Which pathway database should I use for proteomics enrichment analysis?

KEGG pathways are the most commonly used database and provide curated metabolic and signaling pathways. Gene Ontology provides structured vocabularies for biological processes, molecular functions, and cellular components. Reactome provides detailed human pathway annotations with reaction steps. Run enrichment against at least two databases and examine whether pathway findings are consistent across databases. Record the database version because pathway definitions change between releases.

How do I choose the background protein list for over-representation analysis?

The background should be all proteins quantified in your experiment, not the entire proteome. Using the full proteome as background inflates enrichment significance because the background is larger than the detectable protein space. Proteins that were not detected in your experiment cannot contribute to enrichment, so including them in the background distorts the expected counts.

Why do different enrichment tools give different results for the same data?

Tools differ in their statistical tests, multiple testing correction methods, database versions, identifier mapping approaches, and handling of duplicate or missing annotations. A study examining the impact of statistical choices on proteomics functional enrichment found that different statistical approaches produced divergent pathway lists from the same data. Run enrichment with at least two tools and compare results to identify robust pathway findings.

Can I run pathway enrichment on a small proteomics dataset?

Small datasets with limited protein identifications produce unreliable enrichment results because pathway membership counts are low and statistical power is limited. If fewer than a few hundred proteins are quantified, consider whether enrichment analysis is appropriate. Rank-based methods may perform better than over-representation analysis with small datasets because they use the full distribution of measurements instead of a binary threshold.

How do I interpret the direction of pathway enrichment?

Examine whether the proteins from your dataset within an enriched pathway are predominantly increased or decreased in abundance. A pathway enriched with upregulated proteins suggests activation, while enrichment with downregulated proteins suggests suppression. Rank-based methods provide direction information through the enrichment score sign. For over-representation analysis, you must examine the direction of change for pathway member proteins separately.

What should I do when enrichment results conflict with known biology?

Conflicting results may indicate data quality problems, inappropriate statistical methods, database annotation errors, or genuine novel biology. First verify that the analysis parameters are correct and that the background list is appropriate. Check whether the conflicting pathway contains proteins with high missingness or extreme outliers. If the result persists, consider whether the pathway database annotation is current and whether the biological context of your experiment differs from the context in which the pathway was characterized.

How should I report enrichment analysis methods in publications?

Report the software and version, the pathway database and version, the identifier mapping approach, the background protein list definition, the statistical test, the multiple testing correction method, and the significance thresholds. Include the complete enrichment output tables as supplementary data. This documentation enables other researchers to replicate the analysis and assess the robustness of the findings.

Related Bioinformatics Guides

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

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.
• Proteomic signature of dementia risk in type 2 diabetes.. Journal of advanced research, 2026.
• Proteome-Wide Mendelian Randomization Identifies Therapeutic Targets for Abdominal Aortic Aneurysm.. Journal of the American Heart Association, 2025.
• urPTMdb/TeaProt: Upstream and Downstream Proteomics Analysis.. Journal of proteome research, 2023.
• Association Analysis of the Circulating Proteome With Sarcopenia-Related Traits Reveals Potential Drug Targets for Sarcopenia.. Journal of cachexia, sarcopenia and muscle, 2025.
• Plasma Proteomic of Adolescent Idiopathic Scoliosis.. Molecular & cellular proteomics : MCP, 2026.
• Quantitative proteomics analysis of radiation-induced rectal injury in rectal cancer patients undergoing radiotherapy.. Journal of proteomics, 2025.
• Deep Proteomics Analysis Unravels the Molecular Signatures of Tonsillar B Cells in PFAPA and OSAS in the Pediatric Population.. International journal of molecular sciences, 2025.
• Multi-omics analysis of SFTS virus infection in Rhipicephalus microplus cells reveals antiviral tick factors.. Nature communications, 2025.
• AI and Machine Learning for Proteomics-Driven Drug Discovery: Methods, Tools, and Best Practices.. 2026.
• Enabling Next-Generation Mass Spectrometry-Based Proteomics: Standards, Proteoform Resolution, and FAIR, Reproducible, and Quantitative Analysis.. 2026.
• Plasma protein signatures altered before and after tuberculosis diagnosis in a population-based cohort.. 2026.
• The molecular ICU: a primer on omics, informatics and the future of precision critical care.. 2026.
• Causal Disturbance Analysis: A Novel Graph Centrality Based Method for Pathway Enrichment Analysis. IEEE/ACM Transactions on Computational Biology & Bioinformatics, 2020.
• Interpreting the asymmetric interaction between yeast and acetic acid bacteria in kefir grains from a metabolic perspective.. Food Research International, 2026.
• Facilitating pathway and network based analysis of RNA-Seq data with pathlinkR. PLoS Comput. Biol., 2024.
• Integrated enrichment analysis and pathway-centered visualization of metabolomics, proteomics, transcriptomics, and genomics data by using the InCroMAP software. Journal of Chromatography B Analytical Technologies in the Biomedical and Life Sciences, 2014.
• Reproducibility Crossroads: Impact of Statistical Choices on Proteomics Functional Enrichment. International Journal of Molecular Sciences, 2025.
• Meta-analysis of pathway enrichment: Combining independent and dependent omics data sets. Plos One, 2014.

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