Volcano Plots in Metabolomics & Proteomics: Interpretation, Cutoffs, and Best Practices

Volcano plots are a compact way to visualize differential analysis results across thousands of metabolites or proteins. By combining effect size (log₂ fold change) and statistical evidence (−log₁₀ p or q), they surface variables with large, reliable differences between conditions—ideal for screening candidates, generating pathway hypotheses, and communicating results in figures.

What Is a Volcano Plot

A volcano plot is a scatter plot where each point represents a measured feature (e.g., a metabolite or protein) and how its abundance differs between two groups.

• x-axis (log₂ fold change): magnitude and direction of change (right = upregulated; left = downregulated).
• y-axis (−log₁₀ p or q): statistical significance (higher = stronger evidence after testing).

Features with both large effect size and strong statistical support tend to appear in the upper-right (upregulated) or upper-left (downregulated) regions.

How to Interpret: Axes, Cutoffs, and Significant Regions

Locate the center (log₂FC ≈ 0). Variables near zero show little change.

Scan vertically. Higher points indicate stronger statistical evidence (smaller p/q).

Scan horizontally. Points farther from zero indicate larger effect sizes.

Combine both. Upper-left/upper-right areas often contain the most promising candidates.

Check direction. Right = higher in the test group; left = lower.

Common cutoffs (example):

• Effect size: |log₂FC| ≥ 1 (≈ 2-fold)
• Significance: q < 0.05 (Benjamini–Hochberg FDR)

On the plot, draw a horizontal line at −log₁₀(q cutoff) and vertical lines at the log₂FC cutoffs. Predefine thresholds in your analysis plan and keep them consistent across figures.

Key Statistics Behind Volcano Plots

Fold change (FC) and log₂FC: report both direction and magnitude; log scaling centers symmetry around 0.

p-value vs. q-value: use q-values (FDR-adjusted) for ranking and on-figure thresholds to control false positives under multiple testing.

Effect size & confidence: statistical significance alone can be misleading with large n. Consider reporting effect sizes with CIs where appropriate.

Multiple testing: control via Benjamini–Hochberg (or alternatives suited to the design).

Dependencies for meaningful interpretation: adequate normalization, batch correction, appropriate imputation for missing values, and a design with sufficient power.

Customization & Annotation for Better Insight

Color coding: Distinguish significant up/down features (e.g., warm vs. cool tones). For accessibility, consider color-blind-safe palettes and/or shape/opacity cues.

Annotations: Label sentinel metabolites/proteins (top by |log₂FC| and q) and provide tooltips in interactive views.

Consistent styling: Standardize axis labels, legends, title conventions, and threshold lines across figures.

Export options: Prepare both raster (PNG) for web and vector (PDF/SVG) for publication.

Common Pitfalls & QC Checklist

Frequent issues

• Small sample sizes inflate variance and destabilize p/q estimates.
• Inadequate normalization or unaddressed batch effects distort both FC and significance.
• Imputation choices (e.g., constant/LOD imputation) can bias log₂FC—document methods.
• Threshold hacking: altering cutoffs post hoc to “catch” favorites undermines reproducibility.
• Over-reliance on p without considering biological relevance and effect size.
• Small-n and imbalanced designs. With few replicates, variance estimates are unstable and q-values can swing. Use variance-moderating models, assess power where possible, avoid aggressive imputation, and prefer conservative FDR. For imbalanced groups, check leverage and consider precision weights; verify findings with sensitivity analyses and, ideally, external validation.

QC checklist (use before plotting)

• Assess signal drift and apply QC-based normalization if applicable.
• Evaluate batch effects (e.g., PCA trends) and correct (e.g., ComBat, RUV).
• Decide on imputation strategy consistent with data missingness mechanism.
• Pre-register cutoffs and statistical model; document n, covariates, and contrasts.
• Confirm FDR control and report both q-values and effect sizes.

Volcano vs MA vs Manhattan vs VIP–FC

Note: A standard volcano plot uses −log₁₀(p or q) on the y-axis. VIP–FC (PLS-DA) scatter is a separate visualization for model-based variable ranking and should not be labeled a volcano plot.

Applications in Omics: From Screening to Pathway Hypotheses

Volcano plots help triage features into upregulated (upper-right), downregulated (upper-left), and not significant (bottom center), streamlining:

• Biomarker screening and candidate selection
• Pathway hypothesis generation and enrichment (e.g., KEGG, GO)
• Panel building for downstream verification (e.g., PRM/SRM, targeted metabolomics)

Micro-case 1 (Proteomics, Drug Treatment).

A kinase inhibitor vs. DMSO in a cancer cell line (balanced replicates). With normalization and FDR at q<0.05, ~30 phosphoproteins with |log₂FC|>1 cluster in the right arm and enrich MAPK signaling. Five sentinel targets are annotated; follow-up PRM prioritizes four for verification. Pathway analysis suggests MAPK suppression with compensatory PI3K activity.

Figure 1. Volcano plot (proteomics, kinase inhibitor vs DMSO). Points show log₂ fold change vs −log₁₀(q-value). Vertical lines: |log₂FC| = 1; horizontal line: q = 0.05. Sentinel phosphoproteins in the right arm are annotated; candidates enrich MAPK signaling.
 

Micro-case 2 (Metabolomics, Disease vs. Healthy).

Untargeted plasma profiling indicates a left-arm cluster of bile acid conjugates (lower in cases) and a right-arm group of acylcarnitines (higher in cases). A three-metabolite panel reaches cross-validated AUC ≈ 0.87 (external validation recommended). Enrichment points to β-oxidation and bile acid metabolism for follow-up.

Figure 2. Volcano plot (metabolomics, disease vs healthy). Left arm highlights downregulated bile-acid conjugates; right arm highlights upregulated acylcarnitines. Thresholds: |log₂FC| = 1; q = 0.05.

Frequently Asked Questions

Q1: What does a volcano plot show in metabolomics?
A: It visualizes which metabolites differ between groups by combining log₂FC (magnitude/direction) and −log₁₀ p/q (evidence), helping prioritize biologically relevant changes.

Q2: What thresholds should I use?
A: A common starting point is |log₂FC| ≥ 1 and q < 0.05, but adjust based on study goals, sample size, expected effect sizes, and validation plans.

Q3: Should I use p or q on the y-axis?
A: Prefer q-values to reflect multiple-testing control; report both when space allows.

Q4: How does a volcano plot help in proteomics?
A: It highlights proteins with large, statistically supported changes across conditions, accelerating target triage for verification (e.g., PRM/SRM) and pathway analysis.

Q5: What sample size is reasonable?
A: There is no one-size-fits-all. Use power calculations where possible; designs with adequate replicates and batch control yield more stable effect sizes and q-values.

Volcano plots offer an intuitive, statistically grounded lens for discovering meaningful changes in large-scale omics datasets. When coupled with proper normalization, FDR control, and effect-size reporting—and complemented by MA plots, PCA/PLS-DA, and enrichment—they provide a solid bridge from visualization to biological insight.

Try it in Metware Cloud: import data, set FC/q cutoffs, annotate key features, and export publication-ready figures in minutes. Contact our team for guidance on normalization, batch correction, or downstream pathway analysis. 

Read more

• Omics Data Analysis Series
• Metabolomics Batch Effects
• Understanding WGCNA Analysis in Publications
• Deciphering PCA: Unveiling Multivariate Insights in Omics Data Analysis
• Metabolomic Analyses: Comparison of PCA, PLS-DA and OPLS-DA
• WGCNA Explained: Everything You Need to Know
• Harnessing the Power of WGCNA Analysis in Multi-Omics Data
• Beginner for KEGG Pathway Analysis: The Complete Guide
• GSEA Enrichment Analysis: A Quick Guide to Understanding and Applying Gene Set Enrichment Analysis