# Volcano Plots in Proteomics: How to Visualize and Interpret Differential Expression Results Volcano plots are a standard visualization method in quantitative proteomics for displaying differential expression results across thousands of proteins simultaneously. A volcano plot combines statistical significance and effect size on two axes, allowing researchers to quickly identify proteins that change substantially and reproducibly between experimental conditions. This article explains the components of a volcano plot, how to set appropriate thresholds, and how to avoid common misinterpretations related to multiple testing and effect size. The guidance applies to label-free quantification, isobaric tagging approaches such as iTRAQ and TMT, and targeted proteomic assays where differential expression analysis is performed. ## The Purpose of Volcano Plots in Proteomic Analysis Proteomic experiments generate measurements for thousands of proteins in a single run. When comparing two conditions, such as treated versus untreated cells, diseased versus healthy tissue, or samples collected before and after an intervention, researchers need a method to rank which proteins warrant further investigation. A volcano plot serves this purpose by plotting every detected protein on a single graph where the x-axis represents the magnitude of change and the y-axis represents the statistical confidence of that change. The name derives from the visual shape of the plot. Proteins with small changes and low statistical significance cluster near the center, while proteins with large changes and high significance appear at the upper left and upper right corners, creating a shape reminiscent of a volcanic eruption. The practical value of this visualization is that it condenses complex statistical output into a format that supports rapid biological interpretation. In proteomic studies published in recent literature, volcano plots appear consistently as part of standard analysis workflows. For example, a 2025 study of WNT5A signaling in rheumatoid arthritis fibroblast-like synoviocytes used volcano plots alongside heatmaps, enrichment analysis, and protein-protein interaction networks to identify differentially expressed proteins and phosphoproteins after recombinant WNT5A treatment [8]. Similarly, a study examining skin pigmentation changes after kidney transplantation used volcano plots to visualize differentially expressed proteins identified through iTRAQ-based quantitative proteomics [9]. These examples illustrate that volcano plots function as a bridge between raw quantitative data and biological interpretation. ## Components of a Volcano Plot ### The X-Axis: Effect Size The horizontal axis of a volcano plot represents the magnitude of change in protein abundance between two conditions. In most proteomic analyses, this is expressed as the log2 fold change. Using a logarithmic scale serves two purposes. First, it symmetrizes the data so that a doubling of abundance (fold change of 2) and a halving of abundance (fold change of 0.5) appear equidistant from zero. Second, it compresses the wide dynamic range of protein abundances into a manageable visual scale. A log2 fold change of 1 corresponds to a 2-fold increase, a log2 fold change of negative 1 corresponds to a 2-fold decrease, and a log2 fold change of 0 indicates no change. Researchers commonly apply thresholds such as log2 fold change greater than 1 or less than negative 1, which corresponds to a 2-fold change in either direction. Some analyses use more stringent thresholds, such as log2 fold change greater than 2, depending on the biological context and the expected magnitude of protein changes. The choice of effect size threshold should reflect the experimental design and the biological question. In the Alzheimer's disease study using the NULISA targeted proteomic panel, differentially expressed proteins were identified using volcano plots, and specific proteins such as pTau217 showed log2 fold changes of 1.21 in amyloid-positive groups and 4.63 in cognitively impaired groups compared to amyloid-negative controls [7]. These values demonstrate that meaningful biological changes can span a wide range of effect sizes. ### The Y-Axis: Statistical Significance The vertical axis represents the statistical significance of the observed change, typically expressed as the negative base-10 logarithm of the p-value. A p-value of 0.05 corresponds to a negative log10 value of approximately 1.3, while a p-value of 0.001 corresponds to a value of 3. This transformation means that proteins with higher positions on the y-axis have stronger statistical evidence for differential expression. The p-value in proteomic experiments typically comes from statistical tests such as the Student t-test, moderated t-test, or analysis of variance, depending on the experimental design. For experiments with multiple groups, pairwise comparisons are performed and each comparison generates its own volcano plot. The choice of statistical test affects the validity of the p-values and should match the experimental design. ### The Threshold Lines Volcano plots typically include vertical lines marking the effect size threshold and a horizontal line marking the significance threshold. Proteins falling outside both thresholds, in the upper left and upper right regions, are considered significantly differentially expressed. The region between the vertical lines contains proteins that did not change enough in magnitude, regardless of their statistical significance. The region below the horizontal line contains proteins that changed in magnitude but with insufficient statistical confidence. The intersection of these thresholds defines the selection criteria for candidate proteins. A common default is a p-value threshold of 0.05 and a log2 fold change threshold of 1. However, these defaults are not universally appropriate and should be adjusted based on the number of proteins tested, the variability of the measurements, and the biological context. ## Setting Thresholds for Differential Expression ### The Multiple Testing Problem Proteomic experiments test thousands of proteins simultaneously. When performing a statistical test for each protein at a significance level of 0.05, approximately 5 percent of proteins with no true difference will appear significant by chance alone. In an experiment measuring 5,000 proteins, this translates to roughly 250 false positives if no correction is applied. Multiple testing correction methods adjust p-values to control the rate of false discoveries. The Benjamini-Hochberg procedure controls the false discovery rate, which is the expected proportion of false positives among all proteins declared significant. This approach is widely used in proteomics because it balances the need to identify true changes against the risk of reporting false positives. Adjusted p-values, often called q-values, are plotted on the y-axis of volcano plots in place of raw p-values. The PhosPiR pipeline, an automated phosphoproteomic analysis tool, incorporates multiple statistical testing and differential expression analysis as part of its standard workflow, with volcano plots generated as graphical output [10]. This integration reflects the expectation that rigorous proteomic analysis includes appropriate multiple testing correction. ### Effect Size Thresholds and Biological Relevance Statistical significance alone does not indicate biological importance. A protein can show a highly significant p-value with a very small fold change if the measurements are precise and the sample size is adequate. Conversely, a protein with a large fold change may fail to reach statistical significance if the variability between replicates is high. The effect size threshold should be set based on the expected biological magnitude of changes in the system under study. In targeted proteomic assays where the dynamic range is controlled and the assay precision is high, smaller fold changes may be biologically meaningful. In discovery-based mass spectrometry experiments with higher technical variability, larger fold changes provide more confidence that the observed difference reflects biology instead of measurement noise. The kidney transplantation study identified six proteins with significant differential expression out of 2,148 proteins quantified, with KRT7, KRT19, and PTGDS significantly downregulated after transplantation [9]. This relatively small number of significant proteins reflects the use of stringent thresholds appropriate for a small sample size of three paired biopsies. ### Combining Thresholds for Candidate Selection The combination of effect size and significance thresholds defines the criteria for selecting candidate proteins for further investigation. Researchers should document the thresholds used and the rationale for those choices in their analysis methods. The thresholds should be established before examining the results to avoid the bias of adjusting criteria to include or exclude specific proteins. For exploratory analyses, less stringent thresholds may be appropriate to generate hypotheses for validation. For confirmatory analyses, more stringent thresholds reduce the risk of pursuing false leads. The appropriate stringency depends on the stage of the research and the resources available for follow-up experiments. ## Practical Workflow for Creating Volcano Plots ### Step 1: Prepare the Quantitative Data Matrix The input for volcano plot generation is a data matrix with proteins in rows and samples in columns. Each cell contains the quantitative value for a protein in a specific sample. These values may come from label-free quantification, where peptide intensities are compared across runs, or from isobaric labeling approaches such as iTRAQ or TMT, where reporter ion intensities provide relative quantification. Data preprocessing steps include filtering out proteins with missing values, normalizing the data to account for differences in total protein loading between samples, and transforming the data to a logarithmic scale. The specific preprocessing steps depend on the quantification method and the software used for data acquisition. ### Step 2: Define the Comparison Groups Volcano plots require a binary comparison. For experiments with more than two groups, each pairwise comparison requires a separate volcano plot. For example, in a study with control, treatment A, and treatment B groups, three pairwise comparisons are possible: treatment A versus control, treatment B versus control, and treatment A versus treatment B. The comparison groups should be defined based on the biological question. In the Alzheimer's disease study, comparisons were made between amyloid-positive and amyloid-negative groups, tau-positive and tau-negative groups, and cognitively impaired versus unimpaired participants [7]. Each comparison addressed a distinct biological question. ### Step 3: Perform Statistical Testing For each protein, a statistical test compares the mean abundance between the two groups. The choice of test depends on the experimental design. For two-group comparisons with multiple biological replicates, a t-test or moderated t-test is appropriate. For paired designs, such as the before and after kidney transplantation biopsies, a paired test accounts for the relationship between samples from the same individual [9]. The output of this step is a table with columns for protein identifier, log2 fold change, p-value, and adjusted p-value. This table serves as the input for volcano plot generation. ### Step 4: Generate the Plot Volcano plots can be generated using various software tools. The R programming environment with packages from [Bioconductor](https://bioconductor.org/) provides flexible options for creating publication-quality volcano plots. The Bioconductor project maintains documentation for reproducible genomic analysis workflows, including differential expression analysis and visualization [3]. For researchers who prefer graphical interfaces, the [Galaxy Training Network](https://training.galaxyproject.org/) offers accessible workflow training that includes differential expression analysis and visualization components [4]. These platforms provide step-by-step instructions for generating volcano plots without requiring extensive programming experience. ### Step 5: Annotate and Interpret the Results The final step involves annotating the volcano plot with protein names for the most significant candidates and interpreting the results in the biological context. Proteins that meet the significance and effect size thresholds should be listed in a table with their fold changes, p-values, and any relevant annotation information such as gene names or functional categories. The interpretation should consider whether the direction of change is consistent with the biological hypothesis. For example, in the WNT5A study, proteins related to lipid metabolism, cell migration, inflammation, and cell proliferation showed significant changes after treatment, consistent with the known roles of WNT5A signaling in rheumatoid arthritis [8]. ## Software Options and Workflow Considerations ### R and Bioconductor Packages The R environment provides the most flexible options for volcano plot generation. The [Bioconductor](https://bioconductor.org/) project hosts numerous packages for proteomics data analysis, including tools for data import, normalization, statistical testing, and visualization. These packages follow reproducible analysis standards and are documented with extensive vignettes and reference materials [3]. The PhosPiR pipeline demonstrates how multiple R packages can be integrated into a single automated workflow for phosphoproteomic analysis. This pipeline provides data clean-up, statistical testing, differential expression analysis, and graphical output including volcano plots, designed for users without extensive programming knowledge [10]. ### Web-Based Platforms For researchers who prefer point-and-click interfaces, web-based platforms offer accessible alternatives. The [Galaxy Training Network](https://training.galaxyproject.org/) provides tutorials for proteomics data analysis that include volcano plot generation as part of complete analysis workflows [4]. These platforms support reproducibility by recording the analysis steps and parameters used. ### Community Workflow Standards For large-scale or collaborative projects, standardized pipelines offer consistency and reproducibility. The [nf-core documentation](https://nf-co.re/docs) describes community-developed pipeline standards that include quality control, analysis, and reporting components [5]. These pipelines are designed to be portable across computing environments and to produce consistent output formats. ### Choosing Between Options The choice of software depends on several factors: the researcher's programming experience, the scale of the dataset, the need for custom analysis steps, and the requirements of the research community or journal. Researchers should select tools that they can use competently and that produce results that can be documented and reproduced. ## Records and Measurements for Reproducible Analysis ### Documentation Requirements Reproducible volcano plot analysis requires documentation of the complete analysis pathway. This documentation should include the software versions used, the parameters for each analysis step, the thresholds applied, and the rationale for those choices. Version control systems, such as those taught in [The Carpentries lessons](https://carpentries.org/lessons), support tracking changes to analysis scripts and maintaining a record of the analysis history [6]. ### Data Storage and Organization Raw mass spectrometry data files should be stored in their original format and never modified. Processed data tables should be saved in open formats such as CSV or TSV that can be read by multiple software tools. The analysis scripts should be stored alongside the data so that the complete analysis can be rerun if needed. ### Quality Control Records Quality control metrics should be recorded for each mass spectrometry run, including the number of proteins identified, the number of peptides quantified, the coefficient of variation for technical replicates, and the results of any quality control samples included in the experiment. These records support the interpretation of volcano plot results by providing context about data quality. ### Threshold Documentation The thresholds used for defining differentially expressed proteins should be recorded in the methods section of any publication or report. This documentation should include the p-value threshold, whether multiple testing correction was applied and which method was used, and the fold change threshold. Without this documentation, readers cannot assess the stringency of the analysis. ## Common Failure Patterns in Volcano Plot Interpretation ### Ignoring Multiple Testing Correction A frequent error is interpreting raw p-values without correction for multiple testing. In an experiment measuring thousands of proteins, raw p-values produce an unacceptably high number of false positives. Researchers should use adjusted p-values for defining significance and should report which correction method was applied. ### Overemphasizing Statistical Significance Without Effect Size Proteins with very small fold changes can achieve high statistical significance when measurement precision is high. These proteins may not be biologically meaningful even though they pass the significance threshold. The effect size threshold exists to filter out changes that are too small to pursue. ### Overemphasizing Effect Size Without Statistical Significance Conversely, proteins with large fold changes may fail to reach statistical significance when variability is high or sample sizes are small. These proteins may represent true biological changes that the experiment was underpowered to detect, or they may be artifacts of high variability. They should not be reported as significant without additional validation. ### Using Inappropriate Thresholds for the Experimental Design The default thresholds of p-value less than 0.05 and log2 fold change greater than 1 may not be appropriate for all experiments. Studies with small sample sizes, such as the three paired biopsies in the kidney transplantation study, require careful consideration of thresholds to avoid reporting false positives [9]. Studies with highly variable measurements may require more stringent thresholds. ### Misinterpreting the Absence of Significance A protein that does not appear in the significant region of a volcano plot may still be biologically important. The absence of significance could reflect insufficient statistical power, high technical variability, or a true absence of change. Volcano plots show what the experiment detected, not what exists in the biological system. ### Comparing Volcano Plots Across Different Experiments Volcano plots from different experiments cannot be directly compared because the axes scales depend on the specific data distributions, sample sizes, and normalization methods used. A log2 fold change of 1 in one experiment may represent a different biological magnitude than in another experiment with different measurement precision. ## Limitations of Volcano Plot Analysis ### Loss of Information About Individual Samples Volcano plots display summary statistics for each protein, collapsing the individual sample measurements into a fold change and p-value. This summary loses information about the distribution of values across samples, including outliers, bimodal distributions, or batch effects. Researchers should examine the underlying data for candidate proteins before drawing conclusions. ### Inability to Detect Nonlinear Relationships Volcano plots assume that the relationship between conditions is adequately captured by a fold change. This assumption may not hold for proteins with complex regulation patterns, such as those affected by post-translational modifications that alter protein function without changing abundance. Phosphoproteomic analyses, such as those performed by the PhosPiR pipeline, require separate consideration of phosphorylation site changes [10]. ### Dependence on Data Quality The reliability of volcano plot results depends entirely on the quality of the underlying quantitative data. Poorly normalized data, high rates of missing values, or batch effects can produce misleading volcano plots. Quality control should be performed before generating volcano plots, not after. ### Limited Capacity for Multi-Group Comparisons Volcano plots are designed for binary comparisons. Experiments with multiple groups require multiple volcano plots, which can become unwieldy and make it difficult to identify proteins with complex patterns of change across groups. Alternative visualizations, such as heatmaps, may complement volcano plots for multi-group comparisons. ### Sample Size Constraints The statistical power of differential expression analysis depends on the number of biological replicates. Studies with small sample sizes, such as the three paired biopsies in the kidney transplantation study, have limited power to detect small changes and produce p-values with high variability [9]. Researchers should interpret volcano plots from small studies with appropriate caution. ## Quality Controls for Differential Expression Analysis ### Technical Replicates Technical replicates, where the same biological sample is analyzed multiple times, measure the reproducibility of the mass spectrometry workflow. High variability between technical replicates indicates problems with sample preparation, chromatography, or mass spectrometry performance. Technical replicate variability should be assessed before analyzing biological differences. ### Biological Replicates Biological replicates, where independent samples from the same condition are analyzed, capture the natural variability of the biological system. The number of biological replicates determines the statistical power of the differential expression analysis. More replicates provide more reliable p-values and more stable volcano plots. ### Normalization Verification Normalization corrects for systematic differences between samples, such as total protein amount or labeling efficiency. The effectiveness of normalization should be verified by examining the distribution of protein abundances across samples. Poor normalization can create artificial differences that appear as significant changes in volcano plots. ### Batch Effect Assessment Experiments processed in multiple batches can show systematic differences between batches that are unrelated to the biological conditions being compared. Batch effects can produce false positives or mask true differences. Researchers should assess whether batch effects are present and apply appropriate correction methods if needed. ### Positive and Negative Controls Including proteins with known behavior in the experiment provides a check on the analysis. Positive controls should appear as significantly changed in the expected direction. Negative controls should appear unchanged. The absence of expected results indicates a problem with the experimental system or the analysis. ## Professional Escalation Criteria ### When to Seek Specialized Bioinformatics Support Researchers should consider consulting a bioinformatics specialist when the analysis requires advanced statistical methods beyond standard t-tests, when the experimental design is complex with multiple factors or batch structures, or when the results are unexpected and may indicate technical problems. Specialized support can also help when the researcher is unfamiliar with the statistical assumptions underlying the chosen methods. ### When to Repeat the Experiment An experiment may need to be repeated when quality control metrics indicate poor data quality, when the variability between replicates is excessively high, when the results contradict established biological knowledge without a plausible explanation, or when the sample size is too small to support reliable conclusions. Repeating the experiment with additional replicates often provides more value than attempting to extract conclusions from inadequate data. ### When to Validate Candidate Proteins Candidate proteins identified from volcano plots should be validated using an independent method before being reported as definitive findings. Validation methods include targeted mass spectrometry approaches, western blotting, or enzyme-linked immunosorbent assays. The need for validation increases with the importance of the claim and the stringency of the thresholds used. ### When to Reconsider the Experimental Design If the volcano plot shows almost no significant proteins, the experiment may be underpowered, the biological effect may be smaller than expected, or the comparison groups may not differ in the expected way. If the volcano plot shows an implausibly large number of significant proteins, the analysis may have failed to account for batch effects or other systematic biases. Both situations warrant reconsideration of the experimental design. ## Biological Context and Interpretation ### Integrating Volcano Plot Results with Other Analyses Volcano plots provide a list of candidate proteins, but they do not explain the biological significance of those proteins. Interpretation requires integrating the results with functional information. In the WNT5A study, volcano plot results were integrated with enrichment analysis and protein-protein interaction networks to identify key biological processes and signaling pathways regulated by WNT5A, including AMPK, mTOR, VEGFA-VEGFR2, Notch, and endoplasmic reticulum stress pathways [8]. ### Considering Protein Function and Pathway Context The biological interpretation of a differentially expressed protein depends on its known functions and its position in cellular pathways. A protein with a modest fold change may be highly significant if it is a key regulatory node, while a protein with a large fold change may be less important if it is a downstream effector with redundant functions. ### Validating Findings Across Studies Findings from volcano plot analysis should be considered in the context of existing literature. Consistent findings across multiple studies provide stronger evidence than isolated observations. The Alzheimer's disease study noted that the NULISAseq pTau217 assay was comparable to the Simoa pTau217 assay in detecting amyloid PET positivity, with both assays using the ALZpath antibody, providing cross-platform validation of the findings [7]. ### Reporting Effect Sizes and Confidence Intervals Publications reporting volcano plot results should include the effect sizes and confidence intervals for the significant proteins, beyond the p-values. This information allows readers to assess the magnitude of the observed changes and the precision of the estimates. ## Regulatory and Safety Context ### Data Integrity Requirements Proteomic data used in regulatory submissions or clinical decision-making must meet data integrity standards. These standards require that data be attributable, legible, contemporaneous, original, and accurate. The analysis workflow, including volcano plot generation, should be documented and traceable. ### Clinical Translation Considerations When proteomic findings are intended for clinical translation, such as the use of pTau217 for detecting amyloid and tau PET positivity in Alzheimer's disease, additional validation is required [7]. The study reporting these findings noted that the promising pTDP43-409 findings require further replication, illustrating the standard that clinical biomarkers must meet before adoption. ### Reproducibility Standards Funding agencies and journals increasingly require that proteomic analyses be reproducible. This requirement means that the raw data, analysis scripts, and parameters must be available for others to rerun the analysis. Platforms such as [Galaxy](https://training.galaxyproject.org/) and [nf-core](https://nf-co.re/docs) support reproducible analysis by recording workflow steps and parameters [4][5]. ### Data Sharing Requirements Many funding agencies and journals require that proteomic data be deposited in public repositories. The [National Center for Biotechnology Information](https://www.ncbi.nlm.nih.gov/) provides databases for sequence data and related information that support data sharing and reuse [1]. Researchers should plan for data deposition when designing their analysis workflows. ## Training and Skill Development ### Foundational Bioinformatics Skills Researchers who generate and interpret volcano plots benefit from foundational bioinformatics skills, including working with data files, using command-line tools, and understanding basic statistics. The [Carpentries lessons](https://carpentries.org/lessons) provide training in these foundational skills, including shell, Git, and programming concepts [6]. ### Specialized Proteomics Training Specialized training in proteomics data analysis is available through multiple channels. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) program offers learning pathways and practical analysis education for bioinformatics topics, including data resources and analysis methods [2]. The [Galaxy Training Network](https://training.galaxyproject.org/) provides hands-on tutorials for proteomics workflows [4]. ### Continuing Education The field of proteomics data analysis evolves rapidly, with new methods and tools appearing regularly. Researchers should plan for continuing education to stay current with best practices. The [Bioconductor](https://bioconductor.org/) project maintains updated documentation and workflows that reflect current analysis standards [3]. ## At a Glance | Component | Typical Default | Considerations for Adjustment | | --- | --- | --- | | Effect size threshold | Log2 fold change greater than 1 or less than negative 1 | Use larger thresholds for noisy data or small sample sizes, smaller thresholds for targeted assays with high precision | | Significance threshold | P-value less than 0.05 | Apply multiple testing correction for experiments measuring thousands of proteins | | Multiple testing correction | Benjamini-Hochberg false discovery rate | Choose correction method based on the acceptable rate of false positives | | Statistical test | Student t-test or moderated t-test | Use paired tests for paired designs, ANOVA for multi-group comparisons | | Sample size | At least 3 biological replicates per group | More replicates increase power and reliability of p-values | ## Common Analysis Decisions and Their Tradeoffs | Decision | Less Stringent Option | More Stringent Option | Tradeoff | | --- | --- | --- | --- | | Fold change threshold | Log2 fold change greater than 0.58 (1.5-fold) | Log2 fold change greater than 2 (4-fold) | Less stringent captures more candidates but includes more false positives | | P-value threshold | P-value less than 0.05 without correction | Adjusted p-value less than 0.01 | Correction reduces false positives but may miss true small changes | | Missing value handling | Impute missing values | Filter proteins with high missing rates | Imputation can introduce artifacts, filtering reduces coverage | | Normalization method | Global median normalization | Quantile normalization or variance stabilizing normalization | Different methods make different assumptions about the data | ## Decision Framework for Selecting Volcano Plot Thresholds ### The Threshold Selection Problem Researchers often apply default thresholds without considering how those thresholds interact with their specific experimental design. The choice of fold change cutoff and significance level determines which proteins enter the validation pipeline, and the costs of incorrect choices differ depending on the research stage. A discovery experiment designed to generate hypotheses can tolerate more false positives because follow-up validation will filter them out. A confirmatory experiment designed to support a specific claim about a protein's behavior requires fewer false positives because the consequences of reporting an incorrect finding are higher. The decision framework below provides a structured approach for selecting thresholds based on experimental purpose, data quality, and available validation resources. This framework complements the general guidance on threshold setting by making the tradeoffs explicit and actionable. ### Framework Step 1: Classify the Experimental Purpose The first decision is whether the experiment serves a discovery, confirmation, or screening purpose. Discovery experiments aim to identify candidate proteins for further study. These experiments benefit from less stringent thresholds to avoid missing true changes. Confirmation experiments test specific hypotheses about particular proteins. These experiments require more stringent thresholds because the results will be reported as evidence for a specific claim. Screening experiments aim to identify proteins for diagnostic or prognostic applications, such as the targeted proteomic panel used in the Alzheimer's disease study where NULISAseq-pTau217 accurately classified amyloid and tau PET status with ROC areas under the curve of 0.92 and 0.86 respectively [7]. Screening applications require thresholds that balance sensitivity and specificity based on the intended clinical use. ### Framework Step 2: Assess Data Quality and Sample Size The number of biological replicates and the technical variability of the measurements determine the reliability of p-values and fold changes. Experiments with three biological replicates per group, such as the paired skin biopsies in the kidney transplantation study, have limited statistical power and produce p-values with high variability [9]. For such experiments, the fold change threshold should be more stringent because the statistical test has less ability to distinguish true changes from noise. Experiments with six or more replicates per group can support less stringent fold change thresholds because the additional replication provides more stable estimates. The coefficient of variation for technical replicates provides a direct measure of measurement precision. When the coefficient of variation is high, the fold change threshold should be increased to ensure that reported changes exceed the typical measurement noise. When the coefficient of variation is low, smaller fold changes can be interpreted with confidence. ### Framework Step 3: Determine the Acceptable False Discovery Rate The false discovery rate represents the proportion of proteins declared significant that are actually false positives. The acceptable rate depends on the consequences of false positives in the specific research context. For discovery experiments where candidates will be validated, a false discovery rate of 0.10 or 0.15 may be acceptable because the validation step removes false positives. For confirmatory experiments where results will be reported without independent validation, a false discovery rate of 0.01 or 0.05 is more appropriate. The Benjamini-Hochberg procedure controls the false discovery rate and is the standard approach in proteomic analysis. The PhosPiR pipeline incorporates multiple statistical testing as part of its automated workflow, reflecting the expectation that correction is a standard component of rigorous analysis [10]. The chosen false discovery rate should be documented alongside the threshold values. ### Framework Step 4: Consider the Expected Effect Size Distribution The biological system under study influences the range of fold changes that are meaningful. In targeted proteomic assays with high precision, such as the NULISA panel used for Alzheimer's disease research, log2 fold changes of 1.21 for pTau217 in amyloid-positive groups and 4.63 in cognitively impaired groups were reported as biologically meaningful [7]. These values span a wide range, indicating that the threshold should not be set so high that smaller but meaningful changes are missed. For discovery-based mass spectrometry experiments, the dynamic range of protein abundances is wider and technical variability is higher. The expected effect sizes may be larger for highly abundant proteins and smaller for proteins near the detection limit. Researchers should examine the distribution of fold changes across all detected proteins before setting the threshold to ensure that the chosen cutoff captures a reasonable number of candidates. ### Framework Step 5: Match Thresholds to Validation Resources The availability of validation methods affects the optimal threshold stringency. When western blotting, targeted mass spectrometry, or enzyme-linked immunosorbent assays are readily available, less stringent thresholds are acceptable because false positives can be filtered out through validation. When validation resources are limited, more stringent thresholds reduce the number of candidates requiring follow-up and increase the likelihood that each candidate represents a true change. The kidney transplantation study identified six differentially expressed proteins out of 2,148 quantified proteins and validated three of them using immunohistochemical staining [9]. This workflow demonstrates how threshold stringency can be matched to validation capacity. The small number of candidates allowed focused validation efforts on the most promising proteins. ### Framework Step 6: Document and Justify the Threshold Choices The selected thresholds should be recorded with the rationale for each choice. This documentation should include the experimental purpose classification, the sample size and variability assessment, the acceptable false discovery rate, the expected effect size distribution, and the validation resources available. This documentation supports reproducibility and allows readers to assess whether the thresholds were appropriate for the experimental context. The [Bioconductor](https://bioconductor.org/) project provides documentation and workflows that support reproducible analysis practices, including threshold selection and reporting [3]. The [Galaxy Training Network](https://training.galaxyproject.org/) offers tutorials that demonstrate how to document analysis parameters within reproducible workflows [4]. ### Applying the Framework to Common Scenarios For a discovery experiment with five biological replicates per group, moderate technical variability, and access to western blotting for validation, appropriate thresholds might be a false discovery rate of 0.10 and a log2 fold change threshold of 1. This combination captures a broad set of candidates while the validation step filters false positives. For a confirmatory experiment with three biological replicates per group, high technical variability, and no independent validation method available, appropriate thresholds might be a false discovery rate of 0.01 and a log2 fold change threshold of 2. The stringent thresholds compensate for the limited statistical power and the absence of validation. For a screening experiment intended to identify biomarkers, the thresholds should be selected based on the required sensitivity and specificity for the intended clinical application. The Alzheimer's disease study demonstrated that pTau217 could classify amyloid PET positivity with an area under the curve of 0.92, indicating that the threshold selection for screening applications should prioritize the clinical performance characteristics [7]. ### Common Threshold Selection Errors One common error is applying the same thresholds to experiments with different sample sizes. A log2 fold change threshold of 1 may be appropriate for an experiment with ten replicates per group but too lenient for an experiment with three replicates per group. The statistical power difference means that the same threshold produces different false discovery rates in the two experiments. Another common error is selecting thresholds after examining the results to include or exclude specific proteins. This practice introduces bias and produces results that cannot be reproduced. Thresholds should be established before the analysis and documented in the methods. A third error is failing to adjust thresholds when the data quality differs from expectations. If the coefficient of variation is higher than anticipated, the thresholds should be adjusted accordingly. The decision framework provides a structured approach for making these adjustments based on documented criteria instead of ad hoc decisions. ### Recording Threshold Decisions in the Analysis Notebook Each threshold decision should be recorded with the date, the rationale, and the data that informed the decision. This record supports reproducibility and provides a basis for adjusting thresholds in future experiments. The [Carpentries lessons](https://carpentries.org/lessons) provide training in using version control systems to track analysis decisions and maintain a complete record of the analysis history [6]. The [EMBL-EBI Training](https://www.ebi.ac.uk/training) program offers learning pathways that include documentation practices for bioinformatics analyses [2]. ## Frequently Asked Questions ### What is the difference between a volcano plot and a scatter plot of fold changes? A volcano plot adds statistical significance to the fold change information shown in a simple scatter plot. The y-axis displays the negative log of the p-value, so proteins with strong statistical evidence appear higher on the plot. A scatter plot of fold changes alone shows only the magnitude of change and cannot distinguish between changes that are statistically reliable and those that could be due to random variation. ### How do I choose the fold change threshold for my volcano plot? The fold change threshold should reflect the biological context and the technical variability of your measurements. For discovery experiments with high technical variability, a threshold of log2 fold change greater than 1 or 2 provides confidence that observed changes exceed measurement noise. For targeted assays with high precision, smaller thresholds may be appropriate. The threshold should be set before examining the results and documented in the methods. ### Why do I need to correct for multiple testing in proteomics? Proteomic experiments test thousands of proteins simultaneously. At a significance level of 0.05, approximately 5 percent of proteins with no true difference will appear significant by chance. In an experiment measuring 5,000 proteins, this creates roughly 250 false positives without correction. Multiple testing correction methods such as the Benjamini-Hochberg procedure control the proportion of false positives among the proteins declared significant. ### Can I use a volcano plot for experiments with more than two groups? Volcano plots are designed for binary comparisons. For experiments with multiple groups, generate a separate volcano plot for each pairwise comparison of interest. Alternatively, use other visualizations such as heatmaps to display patterns across multiple groups. The choice of comparisons should be guided by the biological questions being addressed. ### What should I do if my volcano plot shows very few significant proteins? Few significant proteins can indicate insufficient statistical power, high technical variability, small biological effects, or a true absence of large changes. Consider whether the sample size is adequate, whether the normalization was effective, and whether the comparison groups actually differ in the expected way. Additional biological replicates often provide more value than adjusting thresholds to force significance. ### How do I validate proteins identified from a volcano plot? Validation requires an independent method that does not rely on the same measurements used for the initial analysis. Options include targeted mass spectrometry, western blotting, or enzyme-linked immunosorbent assays. The validation method should be selected based on the protein of interest and the available resources. Validation is particularly important for proteins that will be reported as definitive findings. ### What information should I report when publishing volcano plot results? Report the software and version used, the normalization method, the statistical test, the multiple testing correction method, the thresholds for significance and effect size, and the number of proteins tested. Include the complete list of significant proteins with their fold changes and adjusted p-values, either in the main text or as supplementary material. ### How do I handle missing values before generating a volcano plot? Missing values in proteomic data can arise from proteins below the detection limit, technical failures, or stochastic sampling in data-dependent acquisition. The handling strategy depends on the missingness pattern. Filtering out proteins with high missing rates reduces the risk of artifacts but decreases coverage. Imputation methods fill in missing values but can introduce bias. The choice should be documented and justified. ## Related Bioinformatics Guides - [Volcano Plot Proteomics: How to Create and Interpret Them Effectively](/knowledge/bioinformatics/volcano-plot-proteomics-how-to-create-and-interpret-them-effectively) - [Proteomics Data Analysis in R: A Practical Workflow for Differential Expression and Visualization](/knowledge/bioinformatics/proteomics-data-analysis-in-r-a-practical-workflow-for-differential-expression-and-visualization) - [How to Interpret Gene Set Enrichment Analysis Results](/knowledge/bioinformatics/how-to-interpret-gene-set-enrichment-analysis-results) - [Spatial Transcriptomics Differential Expression: Methods and Best Practices](/knowledge/bioinformatics/spatial-transcriptomics-differential-expression-methods-and-best-practices) - [RNA-Seq Visualization: Volcano Plots, Heatmaps, and PCA](/knowledge/bioinformatics/rna-seq-visualization-volcano-plots-heatmaps-and-pca) ## 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. - [Targeted proteomic biomarker profiling using NULISA in a cohort enriched with risk for Alzheimer's disease and related dementias.](https://pubmed.ncbi.nlm.nih.gov/40318118). Alzheimer's & dementia : the journal of the Alzheimer's Association, 2025. - [The Role of WNT5A-mediated proteomic and phosphoproteomic regulatory networks in rheumatoid arthritis.](https://pubmed.ncbi.nlm.nih.gov/41237652). Immunobiology, 2025. - [iTRAQ-based quantitative proteomics reveals reduced expression of KRT19, KRT7, and PSTDG in cutaneous specimens after kidney transplantation.](https://pubmed.ncbi.nlm.nih.gov/41006769). Scientific reports, 2025. - [PhosPiR: an automated phosphoproteomic pipeline in R.](https://pubmed.ncbi.nlm.nih.gov/34882763). Briefings in bioinformatics, 2022. - [Phospholipid-Rich DC-Vesicles with Preserved Immune Fingerprints: A Stable and Scalable Platform for Precision Immunotherapy.](https://pubmed.ncbi.nlm.nih.gov/40564018). Biomedicines, 2025. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.