User guide

How to do what Blueprint is for, from a first figure to a survival signature.

Checked against the site on .

Make your first figure

One path from nothing to a downloaded figure: TP53 expression across the tissue types of the TCGA breast cancer cohort.

  1. Create an account: choose a username and a password, press Send code, enter the 6-digit code we email you, then press Create account. You are signed in as soon as the account is created.

  2. Open the Box Plot builder. It is also a card on the Visualization tab of Analysis.

  3. Choose the project TCGA-BRCA and the gene TP53. Keep tissue_type on the x axis and vstd on the y axis. Leave Format on SVG to edit the figure later, or pick PNG or JPEG for slides.

  4. Click Generate. The figure appears as an image you can zoom from 50% to 200% or open full screen.

  5. Download it with the download button in the toolbar. It saves in the format you chose before Generate.

Create an account and sign in

Registration asks for a username (3 to 20 characters: letters, digits, ., _ or -), a password and one contact: an email address, or a China mainland mobile number where SMS sign-up is offered. Press Send code, enter the 6-digit code (valid for 5 minutes by default), then press Create account.

  • Passwords need 8 to 128 characters, including a lowercase letter, an uppercase letter and a number.
  • No code? Check your spam folder, then press Resend.
  • Sign in with your username or email address. 10 failed attempts with one of them, each within 30 minutes of the one before, lock sign-in with it for 15 minutes.
  • Forgot your password? Reset it with an emailed link, valid for 30 minutes, or by SMS where this site offers it.
  • Turn on two-factor authentication in Settings → Account: scan the QR code with an authenticator app, then keep the 10 recovery codes somewhere safe.

Explore TCGA projects

Projects lists the 33 TCGA cancer cohorts. Search by project code, filter by program or disease type, and sort by name, cases, files or date.

  • Search is case-sensitive, so type project codes in capitals: BRCA, not brca.
  • Open a project to see its sample types and its age, gender and race breakdowns.
  • The table of individual samples, with their clinical fields, appears once you sign in.

Make and download a figure

The Visualization tab on Analysis lists every plot type, from box, violin and raincloud plots to correlation matrices, circular heatmaps, radar charts and Sankey diagrams. Each opens the plot builder, which needs you to be signed in.

  1. Choose the data: a TCGA project and gene, or a table you upload. Catalogue plots offer Use Example CSV to show the layout they expect.
  2. Set the options the plot offers, such as the grouping column, colour scheme, statistical test, size and format.
  3. Click Generate. The figure is drawn while you wait; by default a render that takes longer than 30 seconds is stopped.
  4. Download it. Catalogue plots save as SVG, PNG, PDF or JPEG at 50 to 600 DPI; the other plots as PNG, SVG or JPEG.
  • The toolbar zooms from 50% to 200%, shows the figure's details, opens full screen and lists your earlier figures.
  • Charts on the analysis pages are interactive instead: hover over a point to see what it stands for (a sample, a gene or a cell) and its values.

Run a differential expression analysis

Differential expression compares two groups of samples with DESeq2, edgeR, limma-voom or limma.

  1. Choose a TCGA project, or upload a count matrix (CSV, TSV, TXT or Parquet) and a sample sheet (TSV, TXT or CSV).
  2. Pick the metadata column that defines the groups, then the case group and the reference group.
  3. Choose the method and run it. The analysis runs in the background, and the page follows its progress.
  4. Read the results table and the figures: volcano, MA, dispersion (DESeq2 only), p-value histogram, top-gene heatmap and sample distribution.
  • DESeq2, edgeR and limma-voom need raw counts: TPM or FPKM give wrong answers. Use limma for log-scale intensities such as microarray data.
  • Download the table as TSV, CSV or TXT, and each figure as PNG, SVG or PDF.

Run a PCA

PCA runs standard PCA or a variant (incremental, kernel, sparse and mini-batch sparse PCA, truncated SVD, ICA, NMF or LDA) on a TCGA project or your own table (CSV, TSV, TXT, Parquet or h5ad).

  • Hover over a point in the scatter plot to see its sample and coordinates.
  • Export the sample scores as CSV, TSV or Parquet, and the explained variance and feature loadings as CSV or TSV.

Build a survival signature

Survival analysis builds a prognostic gene signature from TCGA overall survival in three stages, on R or Python. Each stage is a background job: Screen starts from a TCGA project, and Signature and Evaluate each start from one of your earlier runs of the stage before.

  1. Screen: a univariate Cox model for each gene, shown as a forest plot, a hazard-ratio bubble plot and Kaplan-Meier curves for the top genes.
  2. Signature: LASSO-Cox with stability selection picks the genes and their coefficients.
  3. Evaluate: split patients into high- and low-risk groups at the median risk score, then check the split with Kaplan-Meier curves, ROC curves, AUC by time horizon and, if you list covariates, a multivariable Cox model.
  • Each gene in the screen has its own page with its survival curve.
  • Kaplan-Meier figures download as PNG, SVG or PDF at 72 to 600 DPI.

Analyse single-cell RNA-seq

Single-cell RNA-seq runs on Scanpy or Seurat in six stages. QC with doublet filtering, reduction and clustering run in order; annotation, trajectory and RNA velocity each start from a clustering, and trajectory and velocity can also start from an annotation.

  • Start from a catalogue dataset, or build your own in the dataset builder from 10x files, a 10x .h5, an .h5ad, a .loom or a text matrix.
  • A dataset can hold up to 50 samples and 150 files, 8 GiB per file and 32 GiB in all, by default.
  • RNA velocity runs on Scanpy only, and its dynamical mode may take up to 45 minutes. The other stages stop after 10 minutes by default.

Use your own data

Upload a table where a page offers it: the plot builder (CSV, TSV or TXT, and XLSX for most plots outside the catalogue), PCA and differential expression. Single-cell data goes through the dataset builder instead.

  • Sign in first. A file uploaded while signed out is stored anonymously, and your analyses cannot use it.
  • The server accepts files up to 50 MiB by default, and some pages check a lower size first: 20 MiB in the plot builder, and 10 MiB for a PCA table or a differential sample sheet.
  • Catalogue plots offer Use Example CSV, and the analysis forms describe the layout they read.

Download TCGA expression files

Downloads lists, for each of the 33 cohorts, the count, TPM and FPKM matrices and the sample information, plus one gene annotation file for all 60,660 genes.

  1. Browse the catalogue. No account is needed.
  2. Sign in and choose a file. Its download link is signed and works for 15 minutes by default.
  3. Check the file against the SHA-256 checksum shown beside it, for example with shasum -a 256.
  • The files are tab-separated. Matrix rows are Ensembl gene ids; match the columns to samples by barcode.

Find earlier work

History lists your figures, from the plot builder and the analysis pages, and your analysis runs, while Keep analysis history is on in Settings → Privacy.

  • Open on a figure takes you to the plot builder for that plot type. Analysis runs are listed for reference; each analysis page offers your earlier runs where a stage needs one.
  • Submitting an analysis with the same settings as an earlier run of yours returns that result instead of running it again. Clear Cache Data on the analysis page forces a fresh run.
  • Figures on the analysis pages render as background jobs, with the same 30-second limit on each render.

Work from scripts

The developer documentation shows how to call the same REST API from curl, Python or R to list projects, run jobs and download results. Access tokens last 15 minutes by default.