feature_group	fields_or_representation	source_and_version	contains_HepG2	contains_CRISPR_or_perturbation_data	encodes_target_function_or_dependency	derived_from_benchmark_outcome	public_before_benchmark	computed_train_only	leakage_risk	major_revision_action	allowed_interpretation
dataset_gene_summary	baseline_mean;baseline_fraction;target_baseline_expression;cell_count summaries	GSE264667 project AnnData snapshot	yes	yes	indirectly	yes_full_dataset_summaries	available_with_dataset	no	moderate_summary_leakage	remove-feature ablation retained the broad result	Sensitivity only; never call fully train-only
depmap_26q1_gene_effect	median_gene_effect;minimum_gene_effect;broad_essential_fraction	DepMap CRISPRGeneEffect 26Q1; SHA256 e610a4cefb13a82b5b256b47eb08b63ff14843f8dbd0fb164bc0a32688e5b89e	yes	yes	yes	no_direct_delta_but_plausible_outcome_proxy	no	no	high_temporal_and_proxy_risk	excluded HepG2 ACH-000739 and reran ridge v1	Outcome-proxy sensitivity only; not temporal validation or leak-free evidence
depmap_26q1_leave_hepg2_out	median_gene_effect;minimum_gene_effect;broad_essential_fraction across 1207 models	DepMap 26Q1 rebuilt excluding ACH-000739	no	yes	yes	no_direct_delta_but_plausible_outcome_proxy	no	no	moderate_to_high_temporal_proxy_risk	five-seed comparison released	Shows single-row robustness only; future release remains a temporal limitation
dmd_integrated_direction_prior	signed source-integrated disease direction	four harmonized DMD source pipelines	not_benchmark_HepG2	no	no	no	mixed_source_dates	no	moderate_source_heterogeneity	renamed; source-specific effects and modules released	Exploratory disease-direction prior, not ground truth
dmd_source_specific_effects	source-median signed effects and source direction counts	four DMD source pipelines	not_benchmark_HepG2	no	no	no	mixed_source_dates	no	low_for_benchmark_leakage_high_for_meta_inference	formal meta-analysis declared not estimable	Source sensitivity only
nmd_scope_annotations	disease gene membership; coverage status	curated sentinel scope matrix	no	no	yes_by_annotation	no	mixed	no	low_model_leakage_high_scope_overreach	reframed as coverage-gap map	Search and coverage only
muscle_expression_context	GTEx median TPM; GSE270868 expression fraction	GTEx v8 and GSE270868 non-injured quadriceps	no	no	no	no	yes	no	low_model_leakage	kept separate from validation	Applicability/expression context only
myoblast_fusion_risk	GSE293514 fusion hit and effect flags	independent human-myoblast CRISPR fusion screen	no	yes	yes_functional_phenotype	no_transcriptomic_delta	public_before_final_candidate_matrix	no	low_for_queue_creation_moderate_for_final_tier	queue predates screen; final matrix adds risk flags	Risk/exclusion filter, not transcriptomic counteralignment or DMD efficacy validation
candidate_observed_delta	observed_rescue_score;response norm;cell support	GSE264667 HepG2 perturbation delta	yes	yes	yes	yes	yes	no	high_if_used_as_prediction_evidence	separated from model validity and labelled observed	Descriptive triage signal only
candidate_model_scores	ridge OOF approximation;MLP LCO;program score	project development outputs	yes	yes	yes	yes_or_derived_from_same_delta	yes	partly	high_circularity_risk	model agreement made an explicit evidence field	Hypothesis ranking only; no independent validation