Observation | NucleoScope

A Worked Example

What does Tail Closed vs. Tail Runaway actually look like?

Four real samples, run through the same pipeline with no manual adjustment: two from TCGA (diagnosed tumors), two from GTEx (an independent healthy-donor tissue reference). Same instrument, same rule, no prior knowledge of which was which.

Runaway cases — diagnosed tumor

TCGA-B3-3925-01Z-00-DX1

Spectrum State🔴 RUNAWAY
RSi P9963.939
Mean RSi4.880
Tail Width (P99–P50)62.010
High-Defect Tail Occupancy (Concavity >10)78.490%
Mean Concavity12.900

TCGA-49-4488-01Z-00-DX10

Spectrum State🔴 RUNAWAY
RSi P9939.265
Mean RSi3.112
Tail Width (P99–P50)37.934
High-Defect Tail Occupancy (Concavity >10)60.540%
Mean Concavity11.271

Closed cases — healthy-donor reference

GTEX-11GSO-0526

Spectrum State🟢 COMPACT
RSi P991.486
Mean RSi0.508
Tail Width (P99–P50)1.222
High-Defect Tail Occupancy (Concavity >10)1.700%
Mean Concavity5.759

GTEX-117XS-2226

Spectrum State🟢 COMPACT
RSi P991.574
Mean RSi0.285
Tail Width (P99–P50)1.374
High-Defect Tail Occupancy (Concavity >10)0.780%
Mean Concavity5.812

The two groups don't overlap, even though samples within a group aren't identical. RSi P99 ranges from 39 to 64 across the two tumor samples, and from 1.49 to 1.57 across the two healthy-tissue samples — a wide spread within the tumor group, a tight cluster within the healthy group, but no overlap between them. All four samples used 10,000 nuclei each.

Four samples, not a statistical result. This illustrates what "Tail Closed" and "Tail Runaway" concretely look like, and that the separation holds even with natural variation between individual samples — it is not, by itself, evidence of a general pattern across a full cohort. Full sample counts and statistics are reported separately; see Validation for how to reproduce this on your own data.

Observation 01

Tail Runaway is a statistical spectrum readout

It is not a single-nucleus label, and not a cancer probability from a trained model — it's the overall shape of the distribution of many nuclei's RSi measurements.

Non-cancer tissue

In the data analyzed so far, mostly presents as Tail Closed

Cancer tissue

In the data analyzed so far, mostly presents as Tail Runaway

Tail status is one readout among several. The Nuclear-State Spectrum also has a peak position, a width, possible multimodal/subpopulation structure, and spatial heterogeneity — cancer is simply the first spectral feature (the tail) with an observed association, not the only thing this instrument can read. Other spectral features may turn out to be informative for questions this report does not address, such as precancerous change, inflammation, aging, drug response, development, or tissue injury and repair — these are unexplored, not claimed.

Observation 02

It recurs in the cancer data personally analyzed so far

As of now, with one known exception, no counterexample has been found in the cancer samples analyzed. This observation currently covers:

Multiple mammalian species (human, rat, dog);
Multiple independent pathology datasets;
33 cancer types analyzed from TCGA.

This describes only the range of data personally analyzed so far. It does not represent an established universal rule, and it has not been confirmed by the broader scientific community.

Observation 03 · Known Boundary

Known exception: frozen sections don't hold

We didn't stop at consistency — we actively looked for where this pattern breaks down. We found one clear exception: TCGA frozen-section samples don't follow the pattern.

Known boundary: Outside of frozen sections, no other exception has been found so far across human, rat, and dog SVS samples and multiple cancer types. This suggests the Tail Runaway / Tail Closed readout may depend on tissue preparation (frozen vs. FFPE), not solely on cancer's structural features — this remains a boundary we're still mapping, and we welcome further independent testing to narrow or overturn it.

Observation 04

The correspondence spans multiple species and tumor types

In the data analyzed so far, the phenomenon appears across different mammalian species and different tumor types (frozen sections excepted).

Appearing across multiple datasets does not mean it has been proven to hold for every cancer, every species, and every experimental condition.