Datasets:

ClarusC64
/

oncology-precancer-constraint-geometry-v0.2

scenario_id string	inflammation_load float64	glucose_variability float64	circadian_disruption float64	immune_surveillance float64	mitochondrial_stress float64	repair_signal_clarity float64	toxin_pressure float64	tissue_turnover_pressure float64	constraint_coupling_score float64	compensatory_capacity_score float64	case_type string	label int64
TR001	0.79	0.43	0.38	0.86	0.46	0.82	0.64	0.61	0.82	0.88	clean	0
TR002	0.62	0.58	0.55	0.47	0.63	0.43	0.5	0.57	0.73	0.45	clean	1
TR003	0.34	0.31	0.57	0.55	0.52	0.18	0.36	0.43	0.54	0.51	unreadability_threshold	1
TR004	0.72	0.69	0.61	0.87	0.58	0.8	0.52	0.64	0.78	0.89	clean	0
TR005	0.49	0.52	0.48	0.44	0.59	0.36	0.47	0.52	0.76	0.39	clean	1
TR006	0.83	0.36	0.3	0.78	0.4	0.74	0.67	0.58	0.69	0.81	clean	0
TR007	0.56	0.64	0.69	0.51	0.72	0.47	0.39	0.5	0.74	0.48	clean	1
TR008	0.68	0.72	0.45	0.84	0.48	0.77	0.42	0.6	0.81	0.86	clean	0
TR009	0.39	0.73	0.74	0.4	0.7	0.42	0.34	0.46	0.63	0.43	clean	1
TR010	0.65	0.4	0.62	0.73	0.43	0.69	0.59	0.53	0.57	0.76	clean	0
TR011	0.45	0.49	0.52	0.58	0.65	0.16	0.45	0.55	0.49	0.54	unreadability_threshold	1
TR012	0.78	0.7	0.63	0.88	0.56	0.83	0.6	0.67	0.84	0.91	clean	0
TR013	0.53	0.57	0.44	0.42	0.68	0.4	0.58	0.62	0.75	0.41	clean	1
TR014	0.85	0.59	0.37	0.81	0.5	0.78	0.63	0.7	0.71	0.84	clean	0
TR015	0.42	0.68	0.8	0.47	0.74	0.39	0.3	0.41	0.66	0.44	clean	1
TR016	0.74	0.46	0.35	0.7	0.55	0.67	0.72	0.65	0.58	0.73	clean	0
TR017	0.59	0.62	0.6	0.49	0.58	0.45	0.52	0.55	0.72	0.46	clean	1
TR018	0.67	0.73	0.58	0.85	0.62	0.76	0.49	0.63	0.79	0.87	clean	0
TR019	0.37	0.42	0.79	0.43	0.71	0.19	0.37	0.49	0.52	0.47	unreadability_threshold	1
TR020	0.8	0.37	0.42	0.83	0.46	0.79	0.7	0.59	0.68	0.83	clean	0
TR021	0.58	0.55	0.5	0.46	0.67	0.37	0.54	0.61	0.77	0.4	clean	1
TR022	0.72	0.65	0.7	0.87	0.59	0.82	0.47	0.66	0.83	0.9	clean	0
TR023	0.43	0.63	0.74	0.48	0.76	0.43	0.33	0.44	0.65	0.45	clean	1
TR024	0.87	0.5	0.34	0.8	0.53	0.73	0.74	0.69	0.74	0.85	clean	0
TR025	0.54	0.6	0.56	0.44	0.64	0.38	0.51	0.58	0.78	0.42	clean	1
TR026	0.76	0.71	0.62	0.89	0.61	0.84	0.58	0.68	0.86	0.92	clean	0
TR027	0.36	0.39	0.83	0.5	0.79	0.17	0.35	0.45	0.55	0.5	unreadability_threshold	1
TR028	0.83	0.56	0.41	0.86	0.47	0.81	0.66	0.71	0.8	0.88	clean	0
TR029	0.5	0.54	0.59	0.45	0.69	0.39	0.56	0.6	0.76	0.41	clean	1
TR030	0.69	0.74	0.53	0.85	0.57	0.77	0.45	0.61	0.7	0.84	clean	0
TR031	0.57	0.52	0.49	0.52	0.61	0.11	0.45	0.54	0.71	0.79	unreadability_threshold	1
TR032	0.69	0.63	0.58	0.94	0.66	0.77	0.61	0.68	0.74	0.48	immune_dominant	0
TR033	0.46	0.59	0.76	0.57	0.72	0.14	0.39	0.47	0.62	0.82	unreadability_threshold	1
TR034	0.73	0.71	0.67	0.91	0.69	0.74	0.66	0.72	0.79	0.52	immune_dominant	0
TR035	0.51	0.48	0.55	0.49	0.58	0.13	0.42	0.5	0.56	0.76	unreadability_threshold	1
TR036	0.77	0.68	0.6	0.93	0.64	0.8	0.7	0.74	0.81	0.55	immune_dominant	0

What this dataset does

This dataset tests whether a model can detect pre-cancer instability from constraint geometry rather than single-variable thresholds.

The task is not cancer diagnosis.

The task is to classify whether a synthetic tissue ecology has crossed into a persistent instability transition.

Core Stability Idea

The dataset represents a stability-transition hypothesis.

Cancer vulnerability may begin when tissue regulation loses self-correcting coherence before visible disease appears.

The positive class does not indicate cancer.

It indicates that the tissue ecology has entered a persistent instability regime.

Prediction Target

label = 1

The tissue-state scenario has crossed into persistent instability transition.

label = 0

The tissue-state scenario remains compensated despite stress.

Row Structure

Each row represents a synthetic tissue-state scenario.

Columns:

scenario_id
inflammation_load
glucose_variability
circadian_disruption
immune_surveillance
mitochondrial_stress
repair_signal_clarity
toxin_pressure
tissue_turnover_pressure
constraint_coupling_score
compensatory_capacity_score
case_type
label

Case Types

clean

Standard stability geometry cases.

Instability emerges through the interaction of coupled pressure and compensation limits.

Examples include:

High coupling plus weak compensation leads to instability.
High coupling plus strong compensation remains stable.

unreadability_threshold

Repair signal clarity falls below a critical readability boundary.

These rows represent the hypothesis that instability may emerge from loss of signal interpretation before visible pathology appears.

In these cases:

Coupling may be moderate.
Compensation may appear strong.
Instability emerges because repair readability has collapsed.

immune_dominant

Immune surveillance remains sufficiently strong to preserve stability despite reduced compensatory capacity.

These rows prevent compensation metrics from becoming deterministic predictors.

In these cases:

Coupling may be high.
Compensation may appear weak.
Stability is maintained through immune dominance.

Anti-Shortcut Design

This dataset contains adversarial boundary cases designed to prevent simple threshold-based solutions.

High constraint coupling does not always imply instability.

High compensatory capacity does not always imply stability.

Low compensatory capacity does not always imply instability.

Low inflammation does not always imply stability.

High inflammation does not always imply instability.

Repair signal clarity can override the apparent reserve profile when it collapses below a critical readability boundary.

Immune surveillance can override apparent reserve weakness when sufficiently strong.

The benchmark is designed so that models must learn interactions rather than individual-variable thresholds.

Evaluation

Submit predictions in the format:

scenario_id,prediction

Run:

python scorer.py predictions.csv data/test.csv

The scorer returns:

accuracy
precision
recall
f1
confusion_matrix
accuracy_clean
accuracy_unreadability_threshold
accuracy_immune_dominant
case_type_accuracy_macro

Scorer Design

The scorer automatically discovers case types from the test set.

It does not rely on a hardcoded list of pathway categories.

Any case type present in the evaluation set automatically receives its own accuracy metric.

This allows the benchmark to evolve without requiring scorer modifications.

Future dataset versions may introduce additional pathway categories such as:

compensation_exhaustion
inflammatory_lock_in
repair_recovery
metabolic_instability
immune_escape

The scorer will automatically generate:

accuracy_

for every case type present in the test set.

This makes the evaluation framework extensible while preserving compatibility across dataset versions.

In addition to overall accuracy, the scorer reports pathway-specific accuracies so researchers can identify which stability-transition mechanisms a model understands and which mechanisms remain failure modes.

Macro Case-Type Accuracy

case_type_accuracy_macro is the unweighted mean of accuracy across case types.

Each pathway receives equal weight regardless of how many rows belong to that pathway.

This prevents dominant case types from masking poor performance on rarer but theoretically important pathways such as unreadability_threshold.

Structural Contribution

Most oncology datasets attempt to predict disease presence.

This dataset attempts to predict stability transition.

The benchmark explicitly represents multiple pathways into instability:

Coupling-driven instability
Unreadability-driven instability
Immune-dominant stabilization

A successful model cannot rely on a single marker.

It must infer the interaction geometry that generates the transition boundary.

Structural Note

This dataset is synthetic.

It is designed to evaluate structural reasoning rather than provide medical diagnosis.

The generator logic is intentionally withheld.

The purpose is to evaluate whether a model can infer instability geometry from interacting constraints rather than relying on simple correlations.

License

MIT

Downloads last month: 22

Number of rows:

Total file size:

15 kB