# Why Are My Enrichment Results Not Significant? Troubleshooting Common Issues in GO and Pathway Analysis ## Key Takeaways - **Input Gene List Size and Coherence:** Insufficient statistical power arises from input lists with fewer than 50 genes, often due to overly stringent differential expression thresholds (e.g., strict adjusted p-value and log2 fold-change cutoffs). Conversely, excessively large input lists dilute signal by approaching the background composition. Biologically heterogeneous lists, mixing distinct cellular responses, also obscure specific enrichment. - **Background Set Definition is Critical:** The background (universe) set must accurately reflect all genes or proteins *measured* and *eligible* for inclusion in the input list, not the entire genome or proteome. Using a broader background inflates expected counts, leading to false negatives by underestimating true enrichment. - **Annotation Quality and Currency:** Sparse or outdated annotations for the organism of study significantly reduce the effective input list size, hindering detection of enrichment. Identifier mismatches (e.g., Entrez vs. Ensembl IDs) and gene symbol ambiguity are common causes of unmapped genes, effectively shrinking the input list. - **Statistical Stringency and Method Choice:** Overly stringent multiple testing correction methods (e.g., Bonferroni) can mask true biological signals. Employing False Discovery Rate (FDR) is generally preferred for balancing sensitivity and specificity. Rank-based methods like Gene Set Enrichment Analysis (GSEA) offer an alternative to Overrepresentation Analysis (ORA) by utilizing the full ranked gene list, mitigating issues related to discrete input list definition and background set selection. - **Data Modality and Specific Considerations:** Proteomics datasets face amplified background issues due to lower detection coverage and dynamic range limitations. Single-cell RNA sequencing data requires accounting for dropout and cell-type heterogeneity, often necessitating aggregation strategies or specialized sparse data methods. --- When gene ontology (GO) and pathway enrichment analyses return no significant terms, the problem usually sits in one of four places: the input gene list, the background set, the annotation database, or the statistical parameters. This article walks through each of these failure points with concrete diagnostic steps and corrective actions. The guidance applies to researchers working with RNA sequencing data, differential expression outputs, proteomics datasets, and other genomics workflows where enrichment testing is a standard downstream step. Enrichment analysis asks a simple question: are certain biological categories overrepresented in your list of interesting genes compared to what you would expect by chance? When the answer comes back negative, the interpretation is often that your biological system has nothing interesting going on. In practice, the negative result frequently reflects a technical or analytical issue that can be identified and corrected. The sections below cover the most common causes of null enrichment results, how to diagnose each one, and what to change in your workflow. ## At a Glance The table below summarizes the most frequent causes of non-significant enrichment results, the diagnostic signs you can check in your own data, and the first corrective action to try. | Common Cause | Diagnostic Sign | First Corrective Action | | --- | --- | --- | | Input gene list too small | Fewer than 50 genes submitted for testing | Relax the differential expression threshold or use rank-based methods like GSEA | | Incorrect background set | Background includes all genes but your measurement platform only detects a subset | Set the background to the set of genes actually measured in your experiment | | Poor or outdated annotation | Genes map to few or no GO terms in the database | Check annotation coverage for your organism and update the annotation package | | Statistical stringency too high | Adjusted p-values all near 1.0 with raw p-values showing signal | Review the multiple testing correction method and consider the false discovery rate | | Redundant or overly broad terms | Significant terms are all parent categories with no specific children | Use an enrichment tool that trims redundant terms or reports term specificity | ## Understanding What Enrichment Analysis Actually Tests Enrichment analysis compares your gene list against a reference set to detect overrepresentation of functional categories. The core logic is straightforward: if you have a list of 500 differentially expressed genes and 50 of them participate in oxidative phosphorylation, the analysis asks whether 50 is more than you would expect given how many oxidative phosphorylation genes exist in the total measured genome. The statistical test itself is only one component of the pipeline. Before any test runs, you have already made several decisions that determine whether the test can produce a meaningful result. These decisions include how you defined your gene list, what you chose as the background, which annotation database you used, and how you handled multiple testing. Each of these decisions can independently produce a null result even when biological enrichment exists. A common misconception is that enrichment analysis is a single tool with fixed behavior. In reality, the term covers a family of methods that differ in their assumptions and inputs. Overrepresentation analysis (ORA) tests a discrete list of genes against a categorical annotation. Gene set enrichment analysis (GSEA) uses ranked gene lists and does not require a significance threshold to define the input. Functional class scoring methods sit between these two approaches. The choice of method should match the structure of your data and the question you are asking. ## Input Gene List Problems The composition and size of your input gene list is the most common source of null enrichment results. Enrichment tests need enough genes in the input list to detect a signal. They also need the list to be biologically coherent, meaning the genes should share some functional relationship that the annotation database can recognize. ### Too Few Genes in the Input List When you submit a list of 10 or 20 genes to an overrepresentation analysis, the statistical power is extremely limited. With a small list, even a strong biological signal may not reach significance because the expected counts in each category are so low. The test simply does not have enough observations to distinguish signal from noise. The typical cause of a small input list is a stringent differential expression threshold. If you called differentially expressed genes with an adjusted p-value below 0.05 and a log2 fold change greater than 2, you may have retained only a few dozen genes. This is especially common in experiments with small sample sizes, high biological variability, or modest effect sizes. Several corrective actions are available. You can relax the fold change threshold while keeping the adjusted p-value threshold. You can use a nominal p-value without multiple testing correction for the initial gene selection, then rely on the enrichment test to control for false discoveries. You can also switch from overrepresentation analysis to a rank-based method like GSEA, which uses the full ranked gene list and does not require a discrete input list. Rank-based methods retain information from genes that fall just below your significance threshold, and this information often carries the enrichment signal. ### Too Many Genes in the Input List The opposite problem also produces null results. If you submit thousands of genes, representing a large fraction of the measured genome, the enrichment test may find that nearly every category is represented at the expected rate. The signal gets diluted because the input list is too close to the background set in composition. This situation arises when the differential expression threshold is too loose or when the biological perturbation affects a very large fraction of the transcriptome. A list of 8,000 genes from a genome of 20,000 measured genes leaves little room for enrichment to be detected. The corrective action is to tighten the gene selection criteria or to use a method that handles broad transcriptional responses. GSEA with a ranked list is often more appropriate for large-scale perturbations because it detects coordinated shifts in gene sets instead of requiring a discrete list of significant genes. Alternatively, you can apply a fold change filter to focus on genes with meaningful effect sizes instead of statistical significance alone. ### Biologically Heterogeneous Input Lists Enrichment analysis assumes that your input genes share some biological relationship. If your gene list mixes multiple independent pathways, each pathway may have too few genes to reach significance. This is common when a perturbation triggers several distinct responses simultaneously, such as inflammation, proliferation, and metabolic remodeling. The solution is to separate the gene list into biologically meaningful subsets before running enrichment. You can cluster genes by expression pattern, by known pathway membership, or by temporal response. Each cluster can then be tested independently. This approach often reveals enrichment that is invisible when all genes are pooled together. ### Technical Artifacts in Gene Selection Genes selected based on technical artifacts instead of biological signal will not show coherent enrichment. Batch effects, dropout in single-cell data, and alignment errors can all produce gene lists that reflect technical variation instead of biology. If your differential expression analysis did not include batch correction and your samples separate by batch in a principal component analysis, the resulting gene list may be dominated by batch-associated genes. Check your quality control metrics before interpreting enrichment results. The Galaxy Training Network provides accessible workflows for RNA sequencing quality control and differential expression analysis that include batch assessment steps. If batch effects are present, rerun the differential expression analysis with batch correction before proceeding to enrichment testing. ## Background Set Errors The background set, also called the universe or reference set, is the list of genes that the enrichment test uses to calculate expected frequencies. An incorrect background set is one of the most common and least recognized causes of null enrichment results. ### What the Background Set Should Be The background should be the set of genes that were actually measured in your experiment and were eligible to appear in your input list. For RNA sequencing, this means all genes with sufficient expression to be tested for differential expression. For microarray data, this means all genes represented on the array. For proteomics, this means all proteins detected in your samples. The background should not be the entire genome if your measurement platform only captures a subset of genes. If you measured 12,000 genes but use all 20,000 protein-coding genes as the background, the expected counts will be inflated. Categories enriched for genes that your platform does not measure will appear depleted, and categories enriched for measured genes will appear less enriched than they actually are. ### How to Diagnose Background Problems A simple diagnostic is to compare the size of your background set to the number of genes measured in your experiment. If the background is substantially larger, you likely have a problem. Another diagnostic is to check whether the enrichment results change dramatically when you alter the background set. If they do, the background is driving the result. Many enrichment tools default to a background of all genes in the annotation database. This default is rarely correct for real experiments. You need to explicitly set the background to the measured gene set. In Bioconductor packages, this is often done by passing the background genes as an argument to the enrichment function. The Bioconductor project provides documentation for reproducible genomic analysis workflows that show how to set the background correctly. ### Tissue-Specific and Condition-Specific Backgrounds In some cases, the background should be restricted further than the measured gene set. If you are studying a tissue that expresses a limited set of genes, using all measured genes as the background may still be too broad. For example, if you are studying liver tissue and your input list contains liver-specific genes, the enrichment test may not find them enriched because those genes are also in the background at their expected frequency. A more appropriate background in this case is the set of genes expressed in the tissue under the control condition. This approach detects enrichment relative to the tissue's baseline expression instead of relative to the whole genome. The tradeoff is that constructing a tissue-specific background requires additional data processing and may reduce the number of genes available for testing. ## Annotation Database Issues Enrichment analysis depends entirely on the quality and completeness of the annotation database. If your genes are poorly annotated, the analysis cannot find enrichment even when it exists. ### Annotation Coverage for Your Organism Model organisms like human, mouse, and yeast have extensive annotation in GO and pathway databases. Non-model organisms often have sparse annotation, with many genes lacking any functional terms. If a large fraction of your input genes have no annotation, the effective input list shrinks dramatically and enrichment becomes difficult to detect. Check the annotation coverage for your organism before running the analysis. The National Center for Biotechnology Information provides data resources for many organisms, and you can check how many of your input genes have GO annotations. If coverage is low, consider using a different annotation source, such as organism-specific databases or orthology-based annotation transfer from a well-annotated relative. ### Outdated Annotation Packages Annotation databases are updated regularly as new experimental evidence accumulates. Using an outdated annotation package can mean that recently characterized genes are missing from the database. This is a particular problem for genes that were functionally characterized in the last few years. Update your annotation packages regularly. Bioconductor releases annotation packages on a schedule, and the Bioconductor project provides installation and update documentation. The EMBL-EBI training materials cover how to work with biological databases and annotation resources in a research context. ### Identifier Mismatches A frequent cause of apparent poor annotation is identifier mismatch. If your gene list uses Entrez IDs but the annotation package expects Ensembl IDs, the analysis will fail to map most of your genes. The result is a very small effective input list and no significant enrichment. Check that your gene identifiers match the annotation database requirements. Most enrichment tools accept multiple identifier types, but you need to specify which type you are using. Conversion tools are available through NCBI and other resources. The Galaxy Training Network provides tutorials that cover identifier conversion as part of RNA sequencing analysis workflows. ### Gene Symbol Ambiguity Gene symbols are not always unique. Some symbols map to multiple genes, and some genes have multiple symbols. If your input list uses gene symbols, ambiguous mappings can cause genes to be dropped or misassigned. This reduces the effective input list and can obscure enrichment. Use stable identifiers such as Entrez IDs or Ensembl IDs instead of gene symbols when possible. If you must use symbols, check for ambiguous mappings and resolve them before running the enrichment analysis. ## Statistical Parameters and Multiple Testing The statistical parameters you choose can determine whether enrichment results reach significance. These parameters include the significance threshold, the multiple testing correction method, and the minimum category size. ### Multiple Testing Correction Methods Enrichment analysis tests thousands of GO terms or pathways simultaneously. Without correction for multiple testing, many terms will appear significant by chance. With overly stringent correction, true enrichment may be missed. The most common correction methods are Bonferroni and the false discovery rate (FDR). Bonferroni controls the family-wise error rate and is very stringent. FDR controls the expected proportion of false positives and is less stringent. For enrichment analysis, FDR is generally preferred because it balances sensitivity and specificity. If your results show no significant terms with Bonferroni correction but many terms with raw p-values below 0.05, try FDR correction. The adjusted p-values will be lower than Bonferroni-adjusted values, and you may recover true enrichment. ### Minimum Category Size Many enrichment tools allow you to set a minimum category size, which filters out terms with very few genes. A common default is to exclude categories with fewer than 5 or 10 genes. If your input list is small, this filter can remove the very categories that would show enrichment. Consider lowering the minimum category size if your input list is small. A category with 3 genes where all 3 appear in your input list of 50 genes is highly enriched, but it will be excluded by a minimum size filter of 5. The tradeoff is that small categories are more susceptible to false positives, so interpret results from small categories with caution. ### Maximum Category Size Some tools also allow a maximum category size to exclude very broad terms. Categories with hundreds or thousands of genes are often too general to be informative. However, if your input list is large, broad categories may be the only ones that reach significance. This is not necessarily a problem, but it may indicate that your input list is too heterogeneous to detect specific enrichment. ### The Multiple Testing Burden The number of categories tested affects the multiple testing burden. If you test 20,000 GO terms, the adjusted p-value threshold will be much more stringent than if you test 500 pathway terms. Some tools test both GO and pathway annotations together, which increases the burden. If you are testing a very large annotation set, consider whether you need to test all categories. You can restrict the analysis to specific GO aspects (biological process, cellular component, molecular function) or to specific pathway databases. This reduces the multiple testing burden and may reveal enrichment that was previously hidden. ## Rank-Based Methods as an Alternative When overrepresentation analysis fails to produce significant results, rank-based methods like GSEA offer an alternative that avoids many of the problems described above. GSEA does not require a discrete input gene list. Instead, it takes a ranked list of all measured genes and tests whether predefined gene sets are enriched at the top or bottom of the ranking. ### Advantages of Rank-Based Methods Rank-based methods retain information from genes that do not reach the significance threshold. A gene with a modest fold change and a p-value of 0.08 still contributes to the ranking. If many genes in a pathway show modest but coordinated changes, GSEA can detect this even when no individual gene reaches significance. GSEA also avoids the background set problem because it uses all measured genes as the ranking basis. There is no need to define a discrete input list or a separate background set. This makes GSEA more robust to the analytical decisions that often cause null results in overrepresentation analysis. ### Implementing GSEA GSEA requires a ranked gene list and a collection of gene sets. The ranking is typically based on a statistic from the differential expression analysis, such as the signed log p-value or the fold change. The gene sets can come from GO, KEGG, or other collections. The Bioconductor project provides packages for GSEA and related methods, with documentation for reproducible workflows. The Galaxy Training Network also offers tutorials that cover GSEA as part of RNA sequencing analysis. These resources provide practical guidance on implementing rank-based enrichment in your own data. ### Interpreting GSEA Results GSEA results include an enrichment score, a normalized enrichment score, and an adjusted p-value. The normalized enrichment score accounts for gene set size and allows comparison across gene sets. Positive scores indicate enrichment at the top of the ranking, and negative scores indicate enrichment at the bottom. GSEA can produce significant results when overrepresentation analysis does not, but it can also produce null results for the same underlying reasons. Poor annotation, identifier mismatches, and inappropriate gene sets will affect GSEA just as they affect ORA. The advantage is that GSEA is less sensitive to the input list and background decisions that commonly cause null results. ## Proteomics-Specific Considerations Enrichment analysis is not limited to transcriptomics. Proteomics datasets present additional challenges that can produce null enrichment results even when the underlying biology shows clear enrichment. ### Protein Detection Coverage Mass spectrometry-based proteomics detects only a fraction of the proteome, and the detected fraction is biased toward abundant proteins. If your enrichment analysis uses a background of all proteins in the annotation database instead of the detected proteins, the expected counts will be wrong. This is the same background problem described above, but it is more severe in proteomics because detection coverage is lower. Set the background to the set of proteins detected in your experiment. This information is available from your proteomics analysis output. The Journal of Proteome Research has published work on proteomics analysis tools that integrate enrichment testing with protein detection data. ### Post-Translational Modification Enrichment If you are studying post-translational modifications, standard GO and pathway databases may not capture the relevant biology. Dedicated PTM databases provide curated lists of proteins known to be substrates of specific modifications. Enrichment testing against these databases can reveal modification-specific regulation that standard annotation misses. The underrepresented post-translational modification database provides curated lists of proteins reported as substrates of underrepresented modifications. Enrichment of these lists in proteomics datasets can reveal unexpected PTM regulation. This approach has been demonstrated in the analysis of proteomics data from diet-induced tissue remodeling and protein interaction studies. ### Protein Abundance and Dynamic Range Plasma and other complex biological fluids present a wide dynamic range of protein abundances. High-abundant proteins can mask the detection of low-abundant proteins, reducing the effective coverage of the proteome. Enrichment strategies that deplete high-abundant proteins or enrich specific protein classes can improve coverage but introduce their own biases. Comparative studies of plasma enrichment methods have shown that different methods enrich different protein classes. Extracellular vesicle preparations enrich vesicle markers, while other methods preferentially capture lipoproteins or cytokines. The choice of enrichment method affects which proteins are detected and therefore which enrichment results are possible. If your proteomics workflow used a specific enrichment strategy, the detected protein set reflects that strategy's biases. ## Single-Cell and Spatial Data Considerations Single-cell RNA sequencing and spatial transcriptomics present additional challenges for enrichment analysis. The sparse nature of single-cell data and the large number of cells create analytical decisions that can produce null enrichment results. ### Dropout and Sparse Data Single-cell RNA sequencing data contain many zeros due to dropout, where transcripts are not detected even though they are expressed. This sparsity affects differential expression analysis and downstream enrichment. Genes with low detection rates may be excluded from the input list, reducing the effective list size. Imputation methods can address dropout, but they introduce their own assumptions. An alternative is to use methods designed for sparse data. Some enrichment approaches for single-cell data aggregate cells into pseudobulk profiles before testing, which reduces sparsity and improves statistical power. ### Cell Type Heterogeneity Single-cell datasets typically contain multiple cell types. If you run enrichment analysis on differentially expressed genes from a mixed population, the input list may combine genes from different cell types with different biological functions. This heterogeneity dilutes the enrichment signal. Cluster the cells first and run differential expression and enrichment within each cluster. This approach identifies cell-type-specific enrichment that is invisible in the mixed population. The protocol literature includes examples of single-cell analysis workflows that integrate clustering, differential expression, and enrichment testing. ### Choosing the Right Aggregation Level The choice of aggregation level affects enrichment results. Testing enrichment on genes differentially expressed between conditions within a single cell type is more specific than testing on genes from the entire dataset. However, the smaller input list may reduce statistical power. A practical approach is to run enrichment at multiple levels: all cells, each cluster, and condition comparisons within clusters. Compare the results across levels to identify consistent enrichment signals. The Galaxy Training Network provides tutorials on single-cell RNA sequencing analysis that include these steps. ## Workflow Integration and Reproducibility Enrichment analysis is one step in a larger bioinformatics workflow. The decisions made in earlier steps affect the enrichment results, and the reproducibility of the entire workflow depends on documenting those decisions. ### Pipeline Standards Community-developed pipeline standards can help ensure that your enrichment analysis is reproducible and comparable across studies. The nf-core project provides documentation for community pipeline standards, including configuration and usage guidance. These pipelines include quality control, alignment, quantification, and differential expression steps that feed into enrichment analysis. Using a standardized pipeline does not guarantee correct enrichment results, but it does ensure that the upstream processing is consistent and documented. This makes it easier to identify which analytical decisions contributed to a null enrichment result. ### Version Control and Documentation Record the versions of all software and annotation databases used in your analysis. Enrichment results can change when annotation databases are updated, and you need to know which version produced your results. Version control for analysis code is equally important. The Carpentries provides lessons on version control with Git and on reproducible data analysis practices. These skills are essential for documenting the analytical decisions that affect enrichment results. ### Containerization Containerization tools package software and dependencies into reproducible units. Using containers for your enrichment analysis ensures that the software environment is identical across runs and across collaborators. This eliminates a source of variability that can produce different enrichment results from the same input data. ## Common Failure Patterns and Their Fixes The table below summarizes common failure patterns in enrichment analysis, the underlying cause, and the specific fix to apply. | Failure Pattern | Underlying Cause | Fix | | --- | --- | --- | | No significant terms with a small input list | Insufficient statistical power | Relax gene selection thresholds or use GSEA | | No significant terms with a large input list | Signal diluted by too many genes | Tighten gene selection or use rank-based methods | | Results change dramatically with background set | Incorrect background | Set background to measured gene set | | Many genes unmapped in the analysis | Identifier mismatch | Convert identifiers to match annotation database | | Significant terms are all broad parent categories | Input list too heterogeneous | Cluster genes before enrichment testing | | Significant results in one annotation but not another | Annotation database differences | Compare results across GO, KEGG, and other databases | | Results differ between software versions | Annotation or software updates | Document versions and update annotation packages | ## Records and Measurements to Keep Good record keeping is essential for diagnosing null enrichment results and for reproducing successful analyses. The following records should be maintained for every enrichment analysis. ### Input Records Record the exact gene list used as input, including the identifier type and the version of the identifier mapping. Record the thresholds used to define the gene list, including the adjusted p-value threshold and the fold change threshold. Record the number of genes in the input list and the number that mapped to the annotation database. ### Background Records Record the background gene set and how it was constructed. If you used all measured genes, record how the measured gene set was defined. If you used a tissue-specific background, record the criteria for inclusion. ### Annotation Records Record the annotation database and version used for the analysis. Record the date the annotation was downloaded or the package was installed. Record the number of genes in the input list that had at least one annotation term. ### Statistical Records Record the statistical test used, the multiple testing correction method, and the significance threshold. Record the minimum and maximum category sizes if these were set. Record the total number of categories tested. ### Output Records Record the full enrichment results, beyond the significant terms. The full results allow you to examine the distribution of p-values and identify whether the analysis was close to significance. A result with many terms at p-values between 0.05 and 0.1 may indicate that a small adjustment to the analytical parameters would reveal significant enrichment. ## When to Escalate to Professional Support Some enrichment analysis problems require expertise beyond what a typical research laboratory can provide. The following situations warrant escalation to a bioinformatics core facility, a collaborator with specialized expertise, or a commercial bioinformatics service. ### Persistent Null Results Across Multiple Approaches If you have tried multiple gene selection thresholds, both overrepresentation analysis and GSEA, and multiple annotation databases, and you still get no significant enrichment, the problem may be in the upstream analysis. Differential expression results that are dominated by technical artifacts will not produce meaningful enrichment. A bioinformatics specialist can review the upstream analysis for issues that are not obvious from the enrichment results alone. ### Non-Model Organism Annotation Gaps If you are working with a non-model organism with sparse annotation, constructing a useful annotation set may require orthology-based transfer from model organisms or manual curation. This is a specialized task that benefits from professional expertise. The NCBI provides data resources that can support this work, but the analysis requires judgment about orthology relationships and annotation quality. ### Complex Multi-Omics Integration If you are integrating transcriptomics, proteomics, and metabolomics data, the enrichment analysis becomes more complex. Multi-omics integration methods combine data from multiple layers to identify coordinated molecular changes. These methods require specialized expertise to implement and interpret. Published protocols describe approaches for integrating and interpreting multi-omics data using unsupervised and supervised methods. ### Regulatory or Clinical Validation Context If your enrichment results will be used in a regulatory submission or a clinical validation study, the analysis must meet higher standards of documentation and reproducibility. Professional bioinformatics support can ensure that the analysis meets these standards and that the results are defensible in a regulatory context. ## Limitations of Enrichment Analysis Enrichment analysis has inherent limitations that can produce null results even when the analysis is technically correct. Understanding these limitations helps you interpret null results appropriately. ### Annotation Bias Annotation databases are biased toward well-studied genes and pathways. Genes involved in common diseases and standard model organisms have extensive annotation, while genes involved in rare processes or non-model organisms have sparse annotation. This bias means that enrichment analysis can only detect enrichment for categories that are well annotated. ### Category Definition Sensitivity The definition of functional categories affects enrichment results. Different databases define pathways differently, and the same biological process may be split across multiple categories in one database and merged in another. The choice of database can determine whether enrichment is detected. ### Statistical Power Limitations Enrichment analysis has limited statistical power for small input lists and for categories with few genes. Even with perfect annotation and correct background, a list of 20 genes cannot reliably detect enrichment for a category containing 5 genes. The statistical power is simply too low. ### Biological Interpretation Limits A significant enrichment result does not prove that the pathway is biologically active. It only shows that the category is overrepresented in your gene list. The result must be interpreted in the context of your experimental design and validated with additional experiments. Conversely, a null result does not prove that the pathway is inactive. It only shows that the analysis did not detect enrichment. ## Frequently Asked Questions ### Why did my enrichment analysis return no results even though I have clear biological differences between my conditions? Clear biological differences do not guarantee significant enrichment results. The enrichment test may lack statistical power if your input gene list is too small, the background set is incorrect, or the annotation database is incomplete. Start by checking the size of your input list and the annotation coverage for your genes. If both look reasonable, review the background set and the multiple testing correction method. A rank-based method like GSEA may detect enrichment that overrepresentation analysis misses. ### How many genes do I need in my input list for enrichment analysis to work? There is no fixed minimum number, but smaller lists require stronger enrichment to reach significance. A list of 50 genes can produce significant results if the enrichment is strong, while a list of 500 genes may produce null results if the signal is diluted. If your list is small, consider relaxing the gene selection thresholds or using a rank-based method that retains information from all measured genes. ### What is the correct background set for enrichment analysis? The background set should be the genes that were actually measured in your experiment and were eligible to appear in your input list. For RNA sequencing, this is all genes with sufficient expression to be tested. For proteomics, this is all proteins detected in your samples. Using the entire genome as the background when you only measured a subset of genes will produce incorrect expected counts and can hide true enrichment. ### Why do my enrichment results change when I use a different annotation database? Different annotation databases define functional categories differently. GO terms are structured as a hierarchy, while pathway databases like KEGG use curated pathway definitions. The same biological process may be represented differently across databases. If your results are highly sensitive to the database choice, the enrichment signal may be weak or the annotation may be incomplete for your organism. ### Should I use overrepresentation analysis or gene set enrichment analysis? Overrepresentation analysis is appropriate when you have a discrete list of significant genes and you want to test whether specific categories are overrepresented. Gene set enrichment analysis is appropriate when you have a ranked list of all measured genes and you want to detect coordinated changes in gene sets. GSEA is often more powerful because it retains information from genes that do not reach the significance threshold. ### My organism is not a model organism and many genes have no annotation. What can I do? Check the annotation coverage for your organism before running the analysis. If coverage is low, consider using orthology-based annotation transfer from a well-annotated relative. Some databases provide ortholog mappings that can be used to transfer functional annotations. You can also focus on pathway databases that may have better coverage for your organism than GO. ### Why did my proteomics enrichment analysis return no significant results? Proteomics datasets typically detect only a fraction of the proteome, and the detected fraction is biased toward abundant proteins. If your background set includes proteins that were not detected, the expected counts will be incorrect. Set the background to the detected proteins. Also consider whether your proteomics workflow used an enrichment strategy that biases the detected protein set toward specific classes. ### How do I know if my null enrichment result is real or a technical artifact? Examine the distribution of p-values from the enrichment test. If many terms have p-values between 0.05 and 0.1, the analysis may be close to significance and small parameter adjustments could reveal enrichment. If all p-values are near 1.0, the input list or background is likely the problem. Check the annotation coverage for your input genes and verify that the background set matches your measured gene set. ## Related Bioinformatics Guides - [How to Interpret Gene Set Enrichment Analysis Results](/knowledge/bioinformatics/how-to-interpret-gene-set-enrichment-analysis-results) - [Pathway Enrichment Analysis for Proteomics: Tools and Interpretation](/knowledge/bioinformatics/pathway-enrichment-analysis-for-proteomics-tools-and-interpretation) - [Pathway Enrichment Analysis in R: Tools and Visualization for Omics Interpretation](/knowledge/bioinformatics/pathway-enrichment-analysis-in-r-tools-and-visualization-for-omics-interpretation) - [Pathway Enrichment Analysis Online: A Guide to Web-Based Tools for Non-Programmers](/knowledge/bioinformatics/pathway-enrichment-analysis-online-a-guide-to-web-based-tools-for-non-programmers) - [Genomic Data Analysis Tools: A Comparative Guide for Researchers](/knowledge/bioinformatics/genomic-data-analysis-tools-a-comparative-guide-for-researchers) ## Related Clinical & Scientific Guides * [A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data](/knowledge/bioinformatics/a-practical-guide-to-detecting-antimicrobial-resistance-genes-in-shotgun-metagenomic-data) * [Computational Immunology: Modeling the Immune System](/knowledge/bioinformatics/computational-immunology-modeling-the-immune-system) * [How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices](/knowledge/bioinformatics/how-to-set-hard-filters-for-germline-variant-calling-a-practical-guide-to-gatk-best-practices) ## References and Further Reading - [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute. - [Bioconductor](https://bioconductor.org/). Bioconductor Project. - [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project. - [nf-core Documentation](https://nf-co.re/docs). nf-core. - [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries. - [Bioinformatics analysis of differentially expressed genes and pathways in the development of cervical cancer.](https://pubmed.ncbi.nlm.nih.gov/34174849). BMC cancer, 2021. - [urPTMdb/TeaProt: Upstream and Downstream Proteomics Analysis.](https://pubmed.ncbi.nlm.nih.gov/35759515). Journal of proteome research, 2023. - [Heparin-enriched plasma proteome is significantly altered in Alzheimer's disease.](https://pubmed.ncbi.nlm.nih.gov/39380021). Molecular neurodegeneration, 2024. - [Mining the plasma proteome: Evaluation of enrichment methods for depth and reproducibility.](https://pubmed.ncbi.nlm.nih.gov/40829696). Journal of proteomics, 2025. - [A model workflow for microfluidic enrichment and genetic analysis of circulating melanoma cells.](https://pubmed.ncbi.nlm.nih.gov/40316673). Scientific reports, 2025. - [Cerebrospinal Fluid-Derived Extracellular Vesicles: A Proteomic and Transcriptomic Comparative Analysis of Enrichment Protocols.](https://pubmed.ncbi.nlm.nih.gov/40791568). Journal of extracellular biology, 2025. - [Automated multiplexed affinity-based enrichment of peptides for LC-MS/MS plasma proteomics.](https://pubmed.ncbi.nlm.nih.gov/39192483). Proteomics, 2024. - [Refined characterization of circulating tumor DNA through biological feature integration.](https://pubmed.ncbi.nlm.nih.gov/35121756). Scientific reports, 2022. - [Preoxygenation When Standard Approaches Fail: Phenotype-Based Strategies for High-Risk Emergent Intubations.](https://doi.org/10.3390/jcm15072477). 2026. - [Protocol for identifying recirculating thymic regulatory T cells and characterizing the role of Eos in their function using scRNA-seq and TCR-seq.](https://doi.org/10.1016/j.xpro.2026.104620). 2026. - [Protocol for investigating mitochondrial structure, function, and metabolism in human cervical cancer cells.](https://doi.org/10.1016/j.xpro.2025.104202). 2025. - [Limited value of Nanopore adaptive sampling in a long-read metagenomic profiling workflow of clinical sputum samples.](https://doi.org/10.1186/s12920-025-02272-8). 2025. - [Protocol for identifying surface membrane proteins and their associated proteome from mouse cortical neuron cultures by in situ biotinylation.](https://doi.org/10.1016/j.xpro.2026.104418). 2026. - [Protocol for integrating and interpreting multi-omics data combining unsupervised and supervised data integrating approaches.](https://doi.org/10.1016/j.xpro.2026.104534). 2026. - [Direct delivery of assay reagents to extracellular vesicles in liquid biopsies for biomarker analysis.](https://doi.org/10.1038/s41596-025-01317-7). 2026. - [An evaluation of cleaning methods, preservation and specimen stages on trace elements in modern shallow marine ostracod shells of Sinocytheridea impressa and their implications as proxies](https://doi.org/10.1016/J.CHEMGEO.2021.120316). 2021. - [P082 : IMPLEMENTATION OF NEXT GENERATION SEQUENCING (NGS) TECHNOLOGY FOR HLA TESTING: KEY LESSONS LEARNED FROM A MULTI-CENTER ALPHA STUDY](https://doi.org/10.1016/J.HUMIMM.2014.08.144). 2014. - [Association of Clinical Timing with Self-Efficacy Among Student Registered Nurse Anesthetists](https://www.semanticscholar.org/paper/218a2bdbf1e14be0fd12d73b9f7e5992fa0ee836). 2022. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.