RNA-Seq Visualization: Volcano Plots, Heatmaps, and PCA

RNA sequencing produces large, high-dimensional datasets that require careful visualization for meaningful biological interpretation. This article explains how to generate and interpret three core visualization types for RNA-seq data: volcano plots for differential expression results, heatmaps for expression patterns across samples and genes, and principal component analysis (PCA) plots for sample relationships and quality assessment. Each visualization answers a different question about your data, and together they form the foundation of standard RNA-seq reporting and quality control.

What Each Visualization Reveals About Your Data

Volcano plots display differential expression results by plotting statistical significance against magnitude of change. The x-axis shows log2 fold change, representing how much a gene's expression differs between conditions, while the y-axis shows negative log10 of the p-value, so genes with stronger statistical evidence appear higher on the plot. Each point represents one gene, and genes that pass both significance and fold change thresholds appear in the upper left and upper right regions of the plot. These plots help you identify candidate genes for further study and assess the overall distribution of expression changes across the genome.

Heatmaps display expression values as a color grid, with rows typically representing genes and columns representing samples. The color intensity indicates expression level, and hierarchical clustering rearranges rows and columns so that samples with similar expression profiles appear adjacent to each other. This visualization reveals sample groupings, gene expression patterns, and whether biological replicates cluster together as expected. Heatmaps with hierarchical clustering are standard outputs from differential expression analysis tools and are used to examine expression patterns of selected gene sets.

PCA plots reduce the dimensionality of your data to show the major sources of variation among samples. Each sample appears as a point, and the axes represent principal components that capture decreasing amounts of variance in the data. The first principal component captures the largest source of variation, the second captures the next largest, and so on. PCA plots reveal whether samples group by experimental condition, whether batch effects are present, and whether any samples are outliers that may need to be removed from analysis.

Data Inputs and Preparation Requirements

Before generating any visualization, you need properly processed count data. RNA-seq analysis typically begins with raw sequencing reads in FASTQ format, which are aligned to a reference genome and quantified to produce count matrices. Public repositories such as the National Center for Biotechnology Information provide access to raw sequencing data and processed count matrices for many published studies, enabling reanalysis and visualization of existing datasets [2]. The European Bioinformatics Institute offers training materials on RNA-seq analysis workflows that cover the steps from raw data to count matrices [1].

Count data must undergo normalization before visualization. Raw counts are not directly comparable across samples because sequencing depth varies between libraries. Common normalization methods include library size normalization, which scales counts based on total reads per sample, and more sophisticated methods that account for gene length and composition biases. The choice of normalization method affects downstream visualization, particularly for heatmaps where expression values are compared across samples.

For differential expression analysis, you need a design matrix that describes the experimental groups being compared. This includes information about which samples belong to which condition, as well as any covariates such as batch, sex, or age that should be accounted for in the statistical model. Differential expression analysis is a fundamental step in RNA-seq data analysis, and tools such as TCC-GUI provide graphical interfaces for performing this analysis with robust normalization and generating visualization outputs including volcano plots and heatmaps with hierarchical clustering [5].

Generating Volcano Plots

Volcano plots require differential expression results as input. These results typically include a gene identifier, log2 fold change, and p-value for each gene. The log2 fold change represents the magnitude of expression difference between conditions, where positive values indicate upregulation and negative values indicate downregulation. The p-value comes from the statistical test comparing expression between groups, and it is often adjusted for multiple testing using methods such as Benjamini-Hochberg correction.

To generate a volcano plot, you plot log2 fold change on the x-axis and negative log10 of the adjusted p-value on the y-axis. Most plotting tools in R, including ggplot2, can create these plots with a few lines of code. You can add horizontal and vertical lines to mark significance and fold change thresholds, and you can color points based on whether they pass these thresholds. Interactive tools such as OmicsVolcano allow users without programming skills to generate and explore volcano plots, highlighting features of interest in high-throughput biological data [6].

The choice of thresholds affects how many genes appear as significant in your plot. A common approach is to use an adjusted p-value threshold of 0.05 and a log2 fold change threshold of 1, which corresponds to a twofold change in expression. However, these thresholds should be tailored to your experimental context. Studies with small sample sizes may require more lenient thresholds, while studies with large sample sizes may identify many genes at stringent thresholds. The volcano plot makes these threshold choices visible and helps you assess whether your results are driven by a few extreme genes or distributed across many genes.

Interpreting Volcano Plot Patterns

The shape of your volcano plot provides information about your differential expression results. A typical volcano plot has a characteristic V shape, with genes clustered near zero fold change and low significance in the center, and genes spreading outward as fold change and significance increase. The upper regions of the plot contain the most statistically significant genes, and the outer regions contain the genes with the largest expression changes.

Genes that appear in the upper right quadrant are significantly upregulated, while genes in the upper left quadrant are significantly downregulated. The number of genes in each quadrant indicates the balance of upregulation and downregulation in your comparison. Some experiments show a symmetric distribution, while others show a bias toward upregulation or downregulation depending on the biological context.

Asymmetry in the volcano plot can indicate technical issues. If most significant genes show positive fold changes, this may reflect a systematic bias in normalization or a genuine biological response. If the plot shows a diagonal band of points, this may indicate that the statistical test is not properly accounting for variance. Examining the distribution of p-values and fold changes can help you determine whether your results are biologically meaningful or driven by technical artifacts.

Generating Heatmaps with Hierarchical Clustering

Heatmaps display expression values for a selected set of genes across all samples. The first step is to choose which genes to include. Common choices include all differentially expressed genes, the top genes ranked by significance or fold change, or a curated set of genes related to your biological question. Including too many genes makes the heatmap difficult to read, while including too few may miss important patterns.

Expression values are typically transformed before plotting. Log2 transformation is standard because it compresses the dynamic range of expression values and makes the distribution more symmetric. For heatmaps, values are often centered and scaled per gene, so that each row shows the relative expression pattern across samples instead of absolute expression levels. This scaling makes it easier to see patterns of upregulation and downregulation across samples.

Hierarchical clustering groups genes and samples based on similarity of expression profiles. The clustering algorithm calculates distances between all pairs of rows and columns, then iteratively merges the most similar items. The resulting dendrograms appear as trees on the sides of the heatmap, showing the relationships between genes and samples. Samples that cluster together have similar expression profiles, which should correspond to biological replicates if the experiment is well controlled.

The heatmap with hierarchical clustering is a standard output from differential expression analysis tools. TCC-GUI includes heatmap visualization with hierarchical clustering as part of its exploratory analysis tools, allowing users to examine expression patterns of differentially expressed genes [5]. The color scale should be chosen to make expression differences visible, with a diverging color scheme such as blue-white-red being common for centered and scaled data.

Interpreting Heatmap Patterns

The most important pattern to look for in a heatmap is whether biological replicates cluster together. If samples from the same condition form a distinct cluster, this indicates that the biological signal is stronger than the technical noise. If replicates are scattered across the dendrogram, this suggests high variability within groups or batch effects that need to be addressed.

The gene dendrogram reveals groups of genes with similar expression patterns across samples. These co-expressed gene clusters may share biological functions or regulatory mechanisms. Examining the genes within each cluster can generate hypotheses about coordinated biological processes. For example, a cluster of genes upregulated in one condition and downregulated in another may represent a pathway that is activated or repressed.

Heatmaps can also reveal unexpected patterns that warrant investigation. If samples cluster by batch instead of by condition, this indicates a batch effect that should be corrected in the statistical analysis. If a single sample shows a distinct expression pattern, this may be an outlier that should be examined for technical issues. The heatmap provides a visual summary of the data that complements the statistical analysis.

Generating PCA Plots

PCA reduces the dimensionality of the expression matrix to a small number of principal components that capture the major sources of variation. The input is a matrix of expression values with genes as rows and samples as columns. Before PCA, the data should be filtered to remove genes with low expression, and the values should be transformed and scaled appropriately.

The PCA calculation produces principal component scores for each sample and loadings for each gene. The scores indicate where each sample falls along each principal component, and the loadings indicate how much each gene contributes to each component. The first principal component captures the largest amount of variance in the data, the second captures the next largest, and so on. Typically, the first two or three components are plotted to visualize sample relationships.

PCA plots show each sample as a point, with the x-axis representing the first principal component and the y-axis representing the second. The percentage of variance explained by each component is usually shown in the axis labels. Samples that are close together in the PCA plot have similar expression profiles, while samples that are far apart are dissimilar. The plot can be colored by experimental condition, batch, or any other sample metadata to reveal groupings.

PCA is widely used in single-cell RNA-seq analysis to identify cell populations and reduce dimensionality before clustering. In the context of single-cell data, PCA is applied to highly variable genes to capture the major sources of biological variation, and the principal components are then used for downstream analyses such as clustering and visualization [11]. The same principles apply to bulk RNA-seq data, where PCA reveals sample relationships and potential batch effects.

Interpreting PCA Plot Patterns

The primary question to answer with a PCA plot is whether samples group by the experimental variable of interest. If samples from the same condition cluster together and are separated from samples of other conditions along the first principal component, this indicates that the condition explains a large proportion of the variance in your data. This separation supports the validity of your differential expression results.

If samples do not group by condition, several explanations are possible. The biological effect may be small relative to other sources of variation, the sample size may be insufficient to detect the effect, or technical factors such as batch effects may dominate the variance. Examining the PCA plot colored by potential confounding variables can help identify the source of the variation.

Outliers are often visible in PCA plots as points that are far from the main cluster of samples. These outliers may result from sample mislabeling, technical failures, or genuine biological differences. Before removing an outlier, you should investigate the cause. Check the sample preparation records, sequencing quality metrics, and alignment statistics to determine whether the outlier reflects a technical problem or a biological phenomenon.

PCA is also used in integrative analyses of multiple data types. Methods such as jointly-hic compute harmonized projections across contact frequency matrices for Hi-C data, producing principal component scores that capture biologically meaningful information beyond traditional compartment scores [10]. These approaches extend the principles of PCA to other genomic data types.

Workflow Integration and Tool Selection

The three visualization types are typically generated in sequence during RNA-seq analysis. After quality control and alignment, you perform differential expression analysis to obtain the results needed for volcano plots. You then select genes of interest for heatmap visualization. PCA is performed on the full expression matrix early in the analysis to assess sample relationships and quality.

Several software options are available for generating these visualizations. R with Bioconductor packages is the most common environment, offering flexibility and a wide range of visualization options. For users without programming experience, graphical interfaces such as TCC-GUI provide access to differential expression analysis and visualization through a web browser [5]. OmicsVolcano offers interactive volcano plot generation without requiring programming skills [6]. Automated pipelines such as ERAPID integrate differential expression analysis with visualization outputs including volcano plots, MA plots, and heatmaps, and can complete analysis of public datasets on a standard laptop in under an hour [17].

The choice of tool depends on your technical skills, the complexity of your analysis, and your need for reproducibility. Script-based approaches in R provide full control and reproducibility but require programming knowledge. Graphical interfaces are easier to use but may limit customization. Automated pipelines offer a balance, providing standardized analysis with visualization outputs while requiring minimal programming.

Quality Control Checks for Each Visualization

Each visualization type serves as a quality control tool in addition to its role in biological interpretation. Volcano plots reveal the distribution of p-values and fold changes, which should be well behaved in properly analyzed data. The p-value distribution should show a uniform distribution for non-significant genes, with an enrichment of small p-values for truly differentially expressed genes. If the p-value distribution is skewed or shows unexpected patterns, this may indicate problems with the statistical model.

Heatmaps reveal whether replicates cluster together and whether the selected genes show coherent expression patterns. If replicates do not cluster together, this indicates high technical variability or batch effects. The heatmap also shows whether the color scale is appropriate for the range of expression values in your data. If most cells are at the extremes of the color scale, the scaling may need adjustment.

PCA plots are particularly useful for detecting batch effects and outliers. If samples cluster by sequencing batch, library preparation date, or other technical factors instead of by biological condition, this indicates a batch effect that should be addressed. Real-time quality control during nanopore sequencing can assess sample and condition variability and determine the number of identified genes per sample, providing early warning of experimental problems [8].

Common Failure Patterns and Troubleshooting

Several common problems can arise when generating and interpreting these visualizations. One frequent issue is that PCA plots show separation by batch instead of by condition. This indicates that technical variation dominates the biological signal. Options for addressing this include adding batch as a covariate in the statistical model, using batch correction methods, or investigating whether the batch effect is confounded with the biological condition.

Another common problem is that volcano plots show very few significant genes. This may result from small sample sizes, high biological variability, or a weak biological effect. Before concluding that there are no differentially expressed genes, check whether the normalization and statistical methods are appropriate for your data. Consider whether the sample size provides sufficient statistical power to detect the expected effect size.

Heatmaps that show no clear clustering pattern may indicate that the selected genes do not distinguish between conditions. This could mean that the differential expression analysis did not identify informative genes, or that the biological differences are subtle. Try selecting genes based on different criteria, such as the top genes by fold change instead of by significance, or examine a broader set of genes.

Outliers in PCA plots require careful investigation. Check whether the outlier sample has unusual sequencing metrics, such as low total reads, low alignment rate, or high duplication rate. Examine whether the outlier is a biological replicate that may have been mislabeled. If the outlier is determined to be a technical failure, it may be removed from the analysis, but this decision should be documented and justified.

Limitations and Interpretation Boundaries

These visualization methods have inherent limitations that should be considered when interpreting results. Volcano plots show only the results of a single pairwise comparison and do not capture the complexity of multi-group experiments. The choice of thresholds for significance and fold change is arbitrary and affects the interpretation of the plot. Results should be interpreted in the context of the statistical methods used and the experimental design.

Heatmaps show relative expression patterns but do not convey absolute expression levels or the uncertainty in expression estimates. The clustering results depend on the distance metric and clustering algorithm used, and different choices can produce different clusterings. The visual impression of a heatmap can be influenced by the color scale and the ordering of rows and columns, so these choices should be made carefully and reported.

PCA captures only a fraction of the total variance in the data, and the first two components may not reveal all relevant structure. If the biological signal is distributed across many components, the PCA plot may not show clear separation between conditions. Higher-order components should be examined to determine whether they capture biologically meaningful variation.

Transcriptomic differences do not always reflect phenotypic differences. A comparison of benign and malignant canine perianal tumors found large differences in the number of differentially expressed genes and enriched pathways, but the malignant or benign behaviors could not be generally mirrored at the mRNA level [9]. This highlights the importance of validating RNA-seq findings with functional experiments and considering the limitations of transcriptomic data.

Reproducibility and Reporting Standards

Reproducibility is essential for RNA-seq visualization. All steps from raw data to final plots should be documented, including software versions, parameter settings, and analysis scripts. Version control systems help track changes to analysis code. Containerization tools can ensure that the analysis environment is reproducible across different computers.

The FAIR Guiding Principles provide a framework for making research data findable, accessible, interoperable, and reusable [4]. Applying these principles to RNA-seq data and analysis code enables others to reproduce and build upon your work. Data should be deposited in public repositories such as NCBI, and analysis code should be shared through platforms such as GitHub.

When reporting visualization results, include the parameters used to generate each plot. For volcano plots, report the significance threshold and fold change threshold. For heatmaps, report the distance metric, clustering method, and scaling approach. For PCA plots, report the number of genes included, the transformation applied, and the percentage of variance explained by each component.

The National Institutes of Health Genomic Data Sharing Policy establishes expectations for sharing genomic data generated with NIH funding [3]. This policy covers data deposition, access, and sharing timelines. Researchers should be aware of these requirements when planning their analysis and reporting workflows.

Records and Documentation Practices

Maintain detailed records of your analysis decisions and their rationale. This includes the software versions used, the normalization method chosen, the thresholds applied, and any data filtering steps. These records support reproducibility and help you and others understand the basis for your conclusions.

Document any samples that were removed from the analysis and the reasons for removal. This includes outliers identified in PCA plots, samples with poor sequencing metrics, and samples that failed quality control. Transparent documentation of data exclusions is essential for scientific integrity.

Record the parameters used for each visualization so that plots can be regenerated if needed. This includes the gene selection criteria for heatmaps, the principal components plotted, and the color scales used. Storing analysis scripts alongside the data ensures that visualizations can be reproduced exactly.

Professional Escalation Criteria

Some situations warrant consultation with a bioinformatics specialist or statistician. If PCA plots show unexpected sample groupings that cannot be explained by the experimental design, seek expert advice before proceeding with downstream analysis. If differential expression results are highly sensitive to the choice of normalization method or statistical model, this instability should be investigated.

If heatmap clustering produces results that contradict known sample relationships, this may indicate a problem with the data or analysis. Consult with a specialist to determine whether the issue reflects a technical artifact or a genuine biological finding. If quality control metrics indicate problems with sequencing depth, alignment, or gene detection, address these issues before interpreting visualization results.

When working with clinical or diagnostic applications, additional scrutiny is required. Real-time transcriptomic analysis has potential applications in clinical settings, but these applications require rigorous validation and regulatory approval [8]. Consult with appropriate experts and follow relevant regulations when applying RNA-seq visualization to clinical questions.

At a Glance

Practical Implementation Steps

Start by loading your count matrix and sample metadata into your analysis environment. Verify that sample identifiers match between the count matrix and metadata, and check that the experimental groups are correctly assigned. Filter out genes with low expression across all samples, as these genes provide little information and add noise to the analysis.

Perform differential expression analysis using an appropriate method for your data. The choice of method depends on your sample size, experimental design, and whether you need to account for batch effects or other covariates. Tools such as TCC-GUI provide access to differential expression pipelines with robust normalization and are suitable for users who prefer a graphical interface [5].

Generate the volcano plot from your differential expression results. Examine the plot to assess the distribution of significant genes and check for unexpected patterns. Adjust thresholds if needed to focus on the most biologically meaningful genes.

Select genes for the heatmap based on your analysis goals. For a general overview, use the top differentially expressed genes ranked by significance or fold change. For hypothesis-driven analysis, use genes from specific pathways or gene sets of interest. Generate the heatmap with hierarchical clustering and examine whether replicates cluster together.

Generate the PCA plot from the full expression matrix. Examine whether samples group by condition and whether any outliers are present. If batch effects are visible, consider whether batch correction is needed before proceeding with downstream analysis.

Document all analysis steps, parameters, and decisions. Save the scripts used to generate each visualization so that the analysis can be reproduced. Record any samples that were removed and the reasons for removal.

Common Failure Patterns

Frequently Asked Questions

What is the difference between a volcano plot and an MA plot?

A volcano plot displays log2 fold change on the x-axis and statistical significance on the y-axis, allowing you to see both the magnitude and significance of expression changes for every gene. An MA plot displays the log2 fold change against the mean expression level, which helps identify whether expression changes are related to expression magnitude. Volcano plots are better for identifying significant genes, while MA plots are better for assessing the relationship between expression level and fold change.

How do I choose the thresholds for a volcano plot?

The thresholds for significance and fold change should be based on your experimental context and goals. A common starting point is an adjusted p-value of 0.05 and a log2 fold change of 1, representing a twofold change. For studies with small sample sizes, you may need more lenient thresholds. For studies with large sample sizes, you may identify many genes at stringent thresholds, and you may want to focus on genes with larger fold changes.

Why do my samples not cluster by condition in the PCA plot?

Several factors can cause this. The biological effect may be small relative to other sources of variation, technical factors such as batch effects may dominate, or the sample size may be insufficient. Examine the PCA plot colored by potential confounding variables such as batch, sex, or age. Consider whether the normalization method is appropriate and whether batch correction is needed.

What does it mean when a heatmap shows clear clustering of genes but not of samples?

This pattern indicates that groups of genes have coordinated expression patterns across samples, but the samples themselves do not show distinct expression profiles. This can occur when the selected genes do not distinguish between the experimental conditions, or when the biological differences between conditions are subtle. Try selecting genes based on different criteria or examine whether the sample clustering is affected by the distance metric or clustering method.

How many genes should I include in a heatmap?

The number of genes depends on your analysis goals and the readability of the plot. Including too many genes makes the heatmap difficult to interpret, while including too few may miss important patterns. A common approach is to include the top 50 to 100 differentially expressed genes ranked by significance or fold change. For hypothesis-driven analysis, you may include a curated set of genes from specific pathways.

Can I generate these visualizations without programming skills?

Yes, several tools provide graphical interfaces for RNA-seq visualization. TCC-GUI is a Shiny-based application that provides differential expression analysis and visualization tools including volcano plots and heatmaps with hierarchical clustering, requiring only a web browser [5]. OmicsVolcano is an interactive tool for volcano plot visualization that does not require programming skills [6]. Automated pipelines such as ERAPID generate visualization outputs as part of an integrated analysis workflow [17].

What should I do if I find an outlier sample in my PCA plot?

Investigate the cause before deciding whether to remove the outlier. Check sequencing quality metrics such as total reads, alignment rate, and gene detection rate. Review sample preparation records to check for mislabeling or technical issues. If the outlier is determined to be a technical failure, document the decision to remove it and note this in your analysis records.

How do I report the parameters used for my visualizations?

Report the software and version used, the normalization method, the thresholds applied, and any data filtering steps. For heatmaps, report the distance metric, clustering method, and scaling approach. For PCA plots, report the number of genes included and the percentage of variance explained by each component. Storing analysis scripts alongside the data ensures that visualizations can be reproduced exactly.

Related Bioinformatics Guides

• RNA-Seq Differential Expression: DESeq2, edgeR, and limma-voom Frameworks
• Alternative Splicing Analysis from RNA-Seq Data
• Single-Cell RNA-seq Clustering and Cell-Type Annotation Pipelines
• RNA Seq Fastq Example: Structural Analysis and Computational Methodologies in Bioinformatics
• Single-Cell RNA-Seq Normalization: Batch Effect Correction and Dimension Reduction (PCA, t-SNE, UMAP)

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.
• TCC-GUI: a Shiny-based application for differential expression analysis of RNA-Seq count data.. BMC research notes, 2019.
• OmicsVolcano: software for intuitive visualization and interactive exploration of high-throughput biological data.. STAR protocols, 2021.
• RAGER: A user-friendly computational platform for integrated analysis of RNA-Seq and ATAC-seq data.. 2026.
• Real-time transcriptomic profiling in distinct experimental conditions.. 2026.
• Extremes on the benign-malignant tumour spectrum: Distinct transcriptomic landscapes between two common canine perianal neoplasms based on the hallmarks of cancer.. 2026.
• Jointly-hic: joint decomposition of contact frequency maps captures salient features of genome architecture across tissues and development.. 2026.
• Single-cell transcriptomic profiling uncovers key molecular signatures in glioma pathogenesis.. 2026.
• Promoter Hypermethylation and Androgen Receptor Regulation Unveil Dual Oxidase 1 as a Tumor Suppressor in Prostate Adenocarcinoma. 2026.
• Duodenal mucosal RNA-Seq identifies coordinated bile acid-axis transcriptional alterations in food-responsive enteropathy in dogs
• To compare transcriptome profiling (RNA-seq) between two wheat cultivars with different antioxidant activities and to clarify the differences between these two wheat cultivars. Journal of Crop Science and Biotechnology, 2025.
• scBubbletree: computational approach for visualization of single cell RNA-seq data. BMC Bioinformatics, 2024.
• Identifying Key Genes and Therapeutic Targets in Down Syndrome: A Comprehensive Analysis Using RNA-Seq, PPI Networks, and Molecular Docking. International Journal of Current Microbiology and Applied Sciences, 2024.
• ERAPID: an end-to-end RNA-seq analysis pipeline for integrative candidate biomarker discovery with applications to neuropsychiatric disorders.. Methods, 2026.
• Practical Compass of Single-Cell RNA-Seq Analysis. Current Osteoporosis Reports, 2023.
• Design of experimental teaching for tumor transcriptome analysis in interdisciplinary curriculum development. Experimental Technology and Management, 2025.

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