| evidence-luad-1 | Classification and external assignment | Retrospective public-cohort classification. Seven reported LUAD input counts sum to 2090 without participant-level proof of disjointness. NTP confidence filtering and incomplete clinical information affect downstream denominators; external template assessment is not prospective clinical validation. | Reported CRD-related gene entries: 2091; TCGA-LUAD input records: 497; Expression clusters: 3; Named external GEO cohorts: GSE72094; GSE68465; GSE50081; GSE42127; GSE41271; GSE31210; Candidate NMF ranks: 2-7; Iterations stated in Methods: 100; NMF method: lee; NTP confidence handling: FDR <0.05; text says records with FDR >0.05 excluded | Retrospective expression/annotation analysis | Retrospective transcriptomic analysis selected three CRD-expression clusters in TCGA-LUAD and assessed transfer of template features to six GEO cohorts. Retrospective public-cohort classification. Seven reported LUAD input counts sum to 2090 without participant-level proof of disjointness. NTP confidence filtering and incomplete clinical information affect downstream denominators; external template assessment is not prospective clinical validation. | PDF p. 3 / print p. 16, Method: Data acquisition and procession and Development and validation of CRD-related subtypes; PDF p. 4 / print p. 17, Results: Identification and Validation; PDF p. 5 / print p. 18, actual NMF and template-feature panels; official Supplementary Fig. S1 and Table S1. Results say five external cohorts but name six GEO accessions. | KF1 |
|---|
| evidence-luad-2 | Retrospective outcome patterns | Kaplan-Meier comparisons describe outcome associations. C2 generally has more favorable survival and C1 generally less favorable survival, but the GSE42127 plot does not show the same ranking throughout. These are not treatment effects or a calibrated clinical risk rule. | Displayed TCGA log-rank P: 0.0062; Displayed GSE50081 log-rank P: 0.00991 | Retrospective expression/annotation analysis | The clusters had different observed survival patterns, with C2 generally more favorable and C1 generally less favorable in the reported analyses. Kaplan-Meier comparisons describe outcome associations. C2 generally has more favorable survival and C1 generally less favorable survival, but the GSE42127 plot does not show the same ranking throughout. These are not treatment effects or a calibrated clinical risk rule. | PDF p. 4 / print p. 17, Identification and Validation Results; PDF p. 5 / print p. 18, actual TCGA Kaplan-Meier panel C; PDF p. 8 / print p. 21, actual proportions and survival panels A-F, despite the Figure 3 functional-enrichment caption; official Supplementary Fig. S2A (GSE50081). | KF2 |
|---|
| evidence-luad-3 | Expression pathway associations | GSVA/GSEA and pathway annotations are computed from bulk tumor expression. They do not directly measure individual circadian phase, sleep behavior or metabolic flux, and do not establish circadian disruption as the cause of the phenotypes. | C1 expression features: Cell cycle, DNA replication and E2F-related enrichment; C2 expression features: Lipid-metabolic and circadian sleep-wake gene-set enrichment; C3 expression features: Cytokine, chemokine, complement and other immune-related expression enrichment | Retrospective expression/annotation analysis | The inferred functional patterns distinguish proliferation-related C1, lipid-metabolic C2 and immune-associated C3. GSVA/GSEA and pathway annotations are computed from bulk tumor expression. They do not directly measure individual circadian phase, sleep behavior or metabolic flux, and do not establish circadian disruption as the cause of the phenotypes. | PDF p. 4 / print p. 17, Method: Exploring specific biological characteristics; PDF p. 7 / print p. 20, Results: Underlying biological functional processes; PDF p. 6 / print p. 19, actual enrichment heatmap and GSEA plots, despite the Figure 2 proportions/survival caption; official Supplementary Fig. S2B. | KF3 |
|---|
| evidence-luad-4 | Genomic annotation associations | TCGA mutation and CNV annotations are associated with the expression-defined clusters. They do not demonstrate that clock disruption caused TP53 changes, genomic instability, proliferation or immune escape. Genomic-data availability need not equal expression-input n. | Genomic analyses: Somatic mutation, GISTIC-based focal/arm CNV, FGA/FGG/FGL and inferred stemness; Genes emphasized in Results: TP53; TTN | Retrospective expression/annotation analysis | TCGA annotation analyses associated C1 with greater mutation and copy-number-alteration burdens. TCGA mutation and CNV annotations are associated with the expression-defined clusters. They do not demonstrate that clock disruption caused TP53 changes, genomic instability, proliferation or immune escape. Genomic-data availability need not equal expression-input n. | PDF p. 4 / print p. 17, Method: Somatic mutation and copy number variation analysis; PDF p. 7 / print p. 20, Results: Genomic alterations; PDF p. 9 / print p. 22, actual mutation waterfall and CNV panels (Figure 4); official Supplementary Fig. S2C. | KF4 |
|---|
| evidence-luad-5 | Inferred immune landscape | Bulk-expression inference and genomic/immune scores describe associations. Greater infiltration can coexist with suppression. These analyses do not directly enumerate cells, prove immune-escape mechanisms or constitute single-cell validation. | Immune inference: 28 ssGSEA subsets; additional expression-inference methods; Additional immune algorithms: ESTIMATE; TIMER; quanTIseq; MCP-counter; EPIC; xCell; Immune/heterogeneity annotations: IPS, S score, HLA expression and external annotation signatures | Retrospective expression/annotation analysis | C3 showed stronger inferred immune infiltration together with immune-suppressive expression features and higher inferred intratumor heterogeneity. Bulk-expression inference and genomic/immune scores describe associations. Greater infiltration can coexist with suppression. These analyses do not directly enumerate cells, prove immune-escape mechanisms or constitute single-cell validation. | PDF p. 4 / print p. 17, Method: Depicting distinct immune landscape; PDF p. 7 / print p. 20, Results: Immune infiltration and immune escape landscape; PDF p. 10 / print p. 23, actual immune-score, HLA and heterogeneity panels (Figure 5); official Supplementary Fig. S2D and Table S2. | KF5 |
|---|
| evidence-luad-6 | Indirect immunotherapy hypotheses | Checkpoint expression, TIS and SubMap provide indirect response hypotheses. The abstract separately reports 79 reference records, but their split, second cohort identity, tumor composition and participant overlap remain unresolved. No prospective LUAD subtype-guided treatment effect is established. | Immune-checkpoint panel: 27; TIS inflammatory gene count: 18; Immunotherapy reference scope: Two reference cohorts, 79 records as reported in Abstract; specific split unverified; SubMap analysis: Expression similarity to responders; nominal and Bonferroni-adjusted SubMap p values; Separate abstract count: 79 immunotherapy records; not combined with the seven LUAD input counts | Computational hypothesis | Checkpoint-expression, TIS and SubMap results generated an immunotherapy-response hypothesis for C3. Checkpoint expression, TIS and SubMap provide indirect response hypotheses. The abstract separately reports 79 reference records, but their split, second cohort identity, tumor composition and participant overlap remain unresolved. No prospective LUAD subtype-guided treatment effect is established. | PDF p. 2 / print p. 15, Abstract Methods (79-record statement); PDF p. 4 / print p. 17, immunotherapy Methods; PDF p. 7 / print p. 20, Results: The assessment of response to immunotherapy; PDF p. 11 / print p. 24, actual checkpoint, TIS and SubMap panels A-C; PDF p. 13 / print p. 26, Data availability. | KF6 |
|---|
| evidence-luad-7 | Predicted drug-response hypotheses | The reported pRRophetic/CTRP/PRISM drug-response estimates are computational. Their axes say estimated AUC while the text calls the endpoint IC50; the text names prednisone while the supplement names prednisolone. No exact IC50 or clinical drug recommendation is inferred. | Drug-prediction method reported: pRRophetic using CTRP and PRISM sensitivity data; Main displayed drug axes: Estimated AUC of docetaxel; Estimated AUC of dexamethasone; Supplementary displayed drug axes: Estimated AUC of paclitaxel; Estimated AUC of prednisolone | Computational hypothesis | Cell-line-based computational drug-response estimates were associated with subtype and remain treatment hypotheses. The reported pRRophetic/CTRP/PRISM drug-response estimates are computational. Their axes say estimated AUC while the text calls the endpoint IC50; the text names prednisone while the supplement names prednisolone. No exact IC50 or clinical drug recommendation is inferred. | PDF p. 7 / print p. 20, Results: Potential drug development; PDF p. 11 / print p. 24, actual drug-response panels under label D, although the caption assigns drugs to D/E; official Supplementary Fig. S2E-F; PDF pp. 12-13 / print pp. 25-26, Discussion limitations. | KF7 |
|---|
| evidence-luad-8 | Within-TCGA signature comparison | Risk scores were recalculated and candidate signatures compared within the same TCGA data. These best-candidate C-index values are exploratory within-cohort discrimination, not external performance or prospective utility. The Chen EG (2017) attribution remains unresolved because reference 41 does not clearly identify the claimed prognostic model. | Candidate prognostic signatures: 151; Article-reported C1 C-index: 0.741; Article-reported C2 C-index: 0.698; Article-reported C3 C-index: 0.796; Reported model labels: Chen EG (2017) for C1/C2; Ma B (2020) for C3, as attributed by this article; Source-attribution boundary: The Discussion names 21-gene and 16-gene models, but their exact source-to-model mapping is not independently established here; reference 41 is a DGCR5/miR-1180 paper | Exploratory within-cohort comparison | The article reports an exploratory comparison of 151 existing prognostic signatures within TCGA subtypes. Risk scores were recalculated and candidate signatures compared within the same TCGA data. These best-candidate C-index values are exploratory within-cohort discrimination, not external performance or prospective utility. The Chen EG (2017) attribution remains unresolved because reference 41 does not clearly identify the claimed prognostic model. | PDF p. 4 / print p. 17, Method: Personalized management for distinct subtypes; PDF p. 12 / print p. 25, Clinical delicacy management and Discussion; PDF p. 11 / print p. 24, actual C-index comparison under panel E (text/caption refer to F, but no F panel is displayed); PDF p. 15 / print p. 28, references 41-42. | KF8 |
|---|