46 Cancer Classes.
46/46 Tail Runaway.

TCGA-OT was built to distinguish 46 cancer classes.

TCGA-OncoTree (TCGA-OT) is a public pan-cancer slide-level classification task introduced with TITAN. It contains 11,186 H&E FFPE diagnostic whole-slide images from TCGA, classified into 46 OncoTree classes, with at least 50 samples per class.

The spectrum is measured, not seen by eye.

Pathology morphology is visible. The nuclear-state spectrum is a quantitative measurement derived from large ensembles of nuclei.

No cancer-type-specific model was trained for this 46-class readout.

Different spectra. Same observed tail regime.

The detailed RSi distributions differ in peak height, peak position and shape. Yet the high-state tail remains open in every tested class.

Exploratory overlay of RSi distributions from the 46 frozen WSIs. One WSI per TCGA-OT class; 10,000 nuclei per WSI.

A 2D nuclear-state spectral peak map.

Tail Runaway is the shared coarse feature. The spectra themselves are not identical. Peak position and normalized peak height place the 46 representative WSIs at different locations in an exploratory spectral map.

Exploratory 46×1 peak map. This figure is not a cancer-type population validation; repeated WSIs per class are required to estimate within-class spectral territories.

Do not trust our plots. Rebuild them.

The exact 46-slide selection is frozen and traceable to GDC file IDs. The goal is to make the observation independently inspectable from the raw public WSI level down to the per-nucleus measurements.

CSV files are expected directly under /tcga-ot/data/. Each table row links to its matching *_NucleusDecoder.csv file.

46 classes. 46 exact public WSIs.

The selection rule was fixed before the Tail results were known. No NucleoScope/Tail result was used to select a slide, and no slide was replaced after its Tail result was known. The table below uses the exact GDC-resolvable manifest supplied with this experiment. fileciteturn36file0L10-L56

Use your own slide selection.

Public source data.

• TCGA-OT / TITAN: Nature Medicine describes TCGA-OT as a 46-class slide-level pan-cancer classification task containing 11,186 H&E FFPE diagnostic WSIs from TCGA. Nature Medicine paper.
• NCI Genomic Data Commons: TCGA slide images can be viewed and downloaded in their original SVS format, and are accessible through the GDC API. GDC Data Portal information.
• Exact experiment manifest: the 46-class frozen manifest on this page records the exact GDC file IDs and selection rule used for this 46×1 observation.