PCA Plot Generator
Paste a data matrix and get a principal component analysis scores plot with the variance explained by each component, samples colored by group, and a downloadable SVG.
Drag & drop a file or
CSV, TSV, Excel (.xlsx/.xls) - max 2000 rows
Scree plot (variance explained per component)
Compressing Thousands of Variables Into a Readable Plane
Principal component analysis, introduced by Pearson (1901) and formalized by Hotelling (1933), solves a problem every high-dimensional dataset has: you cannot look at a thousand variables at once. PCA finds the directions of greatest variance in the data, which are the eigenvectors of the covariance matrix, and projects each sample onto the first two. The result is a single scatter plot that faithfully preserves as much of the original spread as two dimensions can hold.
The percentages on the axes are not decoration. Each principal component accounts for a share of the total variance equal to its eigenvalue divided by the sum of all eigenvalues, so a plot where PC1 and PC2 together explain most of the variance is a trustworthy summary, while one where they explain little is a flat shadow of a richer structure. This tool reports the variance explained by the leading components so you know how much of the picture you are seeing.
One decision shapes the result: scaling. Because PCA chases variance, a variable on a large numeric scale can dominate the components for no reason but its units, so scaling each variable to unit variance, a correlation-based PCA, is the safe default for mixed-scale data. For expression values already on a comparable log scale, an unscaled covariance-based PCA is sometimes preferred, and this tool lets you switch between them.
In practice a PCA plot is a quality-control and exploration step: it reveals whether samples separate by biology, flags outliers, and exposes batch effects that would otherwise contaminate downstream testing. It comes before formal analysis, not instead of it. To visualize the genes driving the separation, pair it with the heatmap generator and the correlation matrix generator, and for the complete analysis with proper statistics, the bioinformatics analysis service runs it end to end.
A Worked Example: One Axis Captures the Whole Story
Six samples, three controls and three treated, are measured on four features and scaled to unit variance. Running this tool's principal component analysis gives the scores below. Almost all the variation collapses onto a single axis.
| Sample | Group | PC1 | PC2 |
|---|---|---|---|
| Ctrl_1 | Control | +1.81 | −0.03 |
| Ctrl_2 | Control | +1.79 | +0.12 |
| Ctrl_3 | Control | +1.87 | −0.08 |
| Treat_1 | Treatment | −1.89 | −0.06 |
| Treat_2 | Treatment | −1.85 | +0.13 |
| Treat_3 | Treatment | −1.73 | −0.07 |
PC1 variance explained
99.7%
separates control from treatment
PC2 variance explained
0.2%
residual within-group noise
PC1 alone explains 99.7% of the variance and cleanly divides the two groups: every control sits near +1.8 and every treated sample near −1.8, while PC2 is essentially noise at 0.2%. Tight clustering within each group and a wide gap between them along PC1 is the visual signature of a strong, reproducible treatment effect with consistent replicates. One caution: the sign of PC1 is arbitrary, so a rerun could place controls on the left instead of the right without any change in meaning.
Common Mistakes When Reading a PCA Plot
- Ignoring the variance-explained percentages. A clean separation on axes that together explain only a small share of the variance may be an illusion, because most of the structure lives in components you are not seeing. Always read the percentages before trusting the clusters.
- Reading absolute distances between clusters. PCA preserves the directions of greatest variance, not literal distances, and the two axes are usually scaled differently. Interpret which samples group together, not how many centimeters apart two clouds sit.
- Over-interpreting the axis sign. The sign of a principal component is arbitrary, so left versus right or up versus down carries no meaning on its own. Only the grouping and the spread matter.
- Forgetting to scale unlike variables. When features are on different numeric scales, a single large-valued feature dominates the components purely through its units. Scale to unit variance unless the values are already comparable, such as log-transformed expression.
- Missing a batch effect. If samples group by sequencing run or processing date rather than by biology, PC1 is capturing a technical artefact. Catch it here, before differential-expression testing, rather than after.
How to Use This Generator
Paste your matrix
One row per sample with the sample name first, then its feature values, under a header of feature names.
Choose scaling
Scale features to unit variance when they are on different scales, running a correlation-based PCA.
Read the scores plot
Samples project onto PC1 and PC2, colored by inferred group, with the variance each explains.
Export
Download the scores plot as a publication-ready SVG.
Next step
Want a PhD to interpret the structure?
Dimensionality reduction, batch-effect diagnostics, and downstream statistics, handled by a PhD statistician.
Our promise: Free pipeline re-run and figure revisions if reviewers push back.
Timeline
Most projects deliver in under 2 weeks. We confirm an exact date in your quote.
If reviewers push back
If reviewers question the pipeline, parameters, or figures, we re-run the analysis and revise free.
Confidentiality
NDA available on request before any project discussion. Your data, study design, and manuscript stay private either way.
Want a PhD methodologist to handle the whole project?
Get a complete dimensionality-reduction and sample-clustering analysis by a PhD statistician. Free pipeline re-run and figure revisions if reviewers push back. Pay only after you approve your quote.
Frequently Asked Questions
What is a PCA plot?
A PCA plot is a scatter plot of samples in the coordinate system of their first two principal components, the directions that capture the most variance in a high-dimensional dataset. Principal component analysis compresses many correlated variables, such as the expression of thousands of genes, into a few uncorrelated components, and plotting the samples on the first two reveals which samples are similar and whether they separate into groups. It is the standard first look at the overall structure of a dataset.
What does PCA stand for?
PCA stands for principal component analysis. It is a dimensionality-reduction method that re-expresses many correlated variables, such as the expression of thousands of genes, as a smaller set of uncorrelated axes called principal components, ordered by how much variance each one captures. A PCA plot then shows the samples positioned on the first two of those components, which is the most informative two-dimensional summary of a high-dimensional dataset.
What does a PCA plot show?
It shows your samples placed by their first two principal components, so how close two points sit reflects how similar the samples are across every variable measured, not just one. From that layout you read the overall structure of the data: whether samples fall into clusters, which samples are outliers, and how much of the total variation each axis accounts for. In gene expression the key question is whether samples separate by biological condition, a good sign, or by batch, a warning of a technical artefact.
How do you interpret a PCA plot?
Samples that sit close together are similar across the measured variables, and samples that separate along an axis differ most in the direction that axis captures. Clear clustering by condition suggests the biological signal dominates the variation, while samples that group by batch instead of biology warn of a technical artefact. The percentage on each axis tells you how much of the total variance that component explains, so a plot where PC1 explains most of the variance and separates the groups is a strong result.
What are PC1 and PC2 in a PCA plot?
PC1, the first principal component, is the single direction through the data that captures the greatest variance; PC2 is the direction capturing the most remaining variance while being uncorrelated with PC1. They are weighted combinations of the original variables, not any one variable, and together they define the plane onto which the samples are projected. The sign of each axis is arbitrary, so a plot can be mirrored without changing its meaning, and higher components capture progressively less variance and are usually not plotted.
Should you scale the data before PCA?
Usually yes, when the variables are on different scales. Because PCA is driven by variance, a variable measured on a large numeric scale will dominate the components purely because of its units. Scaling each variable to unit variance, which turns the analysis into a correlation-based PCA, prevents this. For gene expression that is already log-transformed and comparable, unscaled covariance-based PCA is sometimes preferred; this tool lets you toggle scaling.
What percentage of variance explained is good?
There is no fixed threshold, but you want the first two or three components to capture a substantial share, often more than half, so the plot faithfully represents the data. When PC1 and PC2 explain only a small fraction, the two-dimensional plot hides most of the structure and you should be cautious about reading clusters from it. The variance-explained percentages reported here tell you how much of the picture the plot is actually showing.
Why use PCA on gene expression data?
Expression datasets have thousands of genes but only a handful of samples, and most genes vary together in a few coordinated ways. PCA reduces that redundancy to a few components, making it easy to spot whether samples separate by treatment, to detect outliers, and to catch batch effects before differential-expression testing. It is a quality-control and exploration step that comes before, not instead of, formal statistical analysis.
Related Analysis Tools
To see the genes behind the sample separation as a clustered matrix, the heatmap generator reorders by similarity, and the correlation matrix generator shows pairwise relationships. To select the significant genes first, use the volcano plot generator. For the full analysis with proper statistics, the bioinformatics analysis service takes raw data to results.
Reviewed by
Dr. Sarah Mitchell
Dr. Sarah Mitchell holds a PhD in Biostatistics from Johns Hopkins Bloomberg School of Public Health and has over 15 years of experience in systematic review methodology and meta-analysis. She has authored or co-authored 40+ peer-reviewed publications in journals including the Journal of Clinical Epidemiology, BMC Medical Research Methodology, and Research Synthesis Methods. A former Cochrane Review Group statistician and current editorial board member of Systematic Reviews, Dr. Mitchell has supervised 200+ evidence synthesis projects across clinical medicine, public health, and social sciences. She reviews all Research Gold tools to ensure statistical accuracy and compliance with Cochrane Handbook and PRISMA 2020 standards.
The Tool Is Free. The Full Analysis of Your Biological Data? We Handle That.
Our PhD statisticians run the complete pipeline: differential expression with multiple-testing correction, survival modelling, dimensionality reduction, and publication-ready figures with a reproducible methods section. Constant pricing, most projects delivered in under two weeks.
Our promise: Free pipeline re-run and figure revisions if reviewers push back.