The table downloads in five minutes.
What comes after is the hard part.
TCGA leaves a clinical table of 12,591 samples, 11,081 patients and 33 cancer types open to anyone — along with mutations, TMB, copy number and expression for 566 lung adenocarcinomas.
These 20 lessons open that table one cell at a time: the twenty-eight characters of a barcode, counts versus TPM, the three kinds of blank, the four survival endpoints, the TMB denominator nobody writes down, the fingerprints left by the hospital that sent the tissue. Python runs right inside this page. There is nothing to install.
How you'll learn
Before any analysis, ask what the column means. The fourth field of the barcode, the unit of an expression value, the three kinds of blank — get these wrong and everything downstream is wrong.
Not a stand-in dataset. The values from the TCGA Pan-Cancer Clinical Data Resource and cBioPortal PanCanAtlas 2018, exactly as published.
You move the cut-off seventeen times and test all nineteen genes. Watching a housekeeping gene come out “significant” is what stops you being fooled later.
Every lesson has a laboratory widget — splitting a barcode, switching units, sliding a cut-off, redrawing a survival curve, all with your own hands.