CNA evidence and diagnostic uncertainty

Stage 05 describes molecular CNA patterns and evidence limits. It does not infer a validated tumor diagnosis from a catalog overlap. Recurrent CNA regions are shared across cancer lineages, and a broad alteration can contain many genes. These findings motivate preserving molecular descriptions rather than selecting a tissue from the first matching rule. Zack et al., Nature Genetics.

Read the evidence fields

The classification table includes:

Field	Meaning
matched_cna_patterns	All matched molecular rules; simultaneous matches remain visible
cna_evidence_status	Detected CNA, catalog overlap, absent CNA or inconsistent inputs
n_distinct_driver_supporting_segments	Distinct chromosome/start/end/state combinations supporting catalog hits
driver_segment_support_status	Whether supporting coordinates are complete, partial or absent
n_shared_driver_supporting_segments	Segments overlapping multiple catalog regions
n_partial_catalog_region_overlaps	Catalog regions with a partial overlap in at least one hit
cna_uncertainty_flags	Shared segments, partial overlap and other interpretation limits

A distinct segment is not necessarily an independent biological event. Overlap counts do not establish focal amplification, gene activation, sequence mutation or biallelic loss. For example, EGFR-region gain alone does not establish an integrated glioma diagnosis; CNS classification uses histology and defined molecular criteria. WHO CNS5 summary.

Broad-cancer mode leaves tissue origin unresolved. A restricted sample set is a supplied study context, explicitly recorded as a prior. It is not a tissue prediction. Pathology assessment requires integration with morphology, immunophenotype and molecular findings. International Consensus Classification.

Scores and probabilities

The evidence score is a hand-weighted research summary. Shared catalog hits are capped by available supporting-segment counts. Retrieved literature supplies background and adds no points. By default, probability fields are empty and marked not_estimated_no_reference_labels.

An optional binary-labelled table can fit a logistic mapping for its matching score target. agreement_score calibration is not reused for probable_cna_score; generic score columns target agreement. Estimates from a user table are labeled as fitted with external validation not established.

Before clinical probability claims, an unchanged model and endpoint require independent validation in the intended population, including calibration, discrimination and uncertainty. Fitting a supplied table alone does not provide that evidence. TRIPOD statement.

The clinician HTML/PDF reports expose these limits alongside the underlying CNA findings. CNA-flat or missing results do not establish absence of cancer.