A genomically validated loss of chromosome Y transcriptional program is a modest prognostic signal independent of tumour stage in male lung adenocarcinoma

Background. Somatic loss of chromosome Y (LOY) is the most common acquired mutation in the male genome and an emerging determinant of cancer aggressiveness and immune evasion, yet whether an LOY-associated transcriptional program constitutes a prognostic axis independent of tumour stage and generic proliferation in non-small-cell lung cancer has not been established. We asked whether LOY defines a genomically-anchored, stage-independent prognostic signal in male lung adenocarcinoma (LUAD). Methods. We analysed TCGA-LUAD bulk RNA-seq (STAR TPM; UCSC Xena) restricted to 231 male primary tumours with overall-survival (OS) follow-up (86 deaths; median follow-up 29.2 months). A male-specific-Y (MSY) transcriptional score classified tumours as LOY (lowest quartile, n = 58) or Y-intact (n = 173). The MSY-based call was benchmarked against genomic chromosome-Y copy number (segment-level SNP6) and against ABSOLUTE tumour purity. Differentially expressed genes (DEGs) were screened by univariate Cox and reduced by L1-penalised (LASSO) Cox; a per-tumour risk score was evaluated by repeated 5-fold cross-validation (CV) with the whole supervised chain re-derived inside each fold. The risk score was tested for independence from TNM stage, age and smoking (multivariable and stage-stratified Cox), from a 19-gene proliferation meta-score, and for gene-selection stability. External validation used the male subset of GSE31210. Results. The transcriptional LOY call agreed substantially with genomic Y loss (Cohen's κ = 0.72; MSY score vs chrY copy number Spearman ρ = 0.84, p = 2×10⁻⁶¹), confirming that low MSY expression reflects genuine genomic LOY rather than an artefact of tumour purity (purity did not differ between groups, p = 0.33; MSY score was uncorrelated with purity, ρ=−0.02). DEGs were robust to a non-parametric test (370/371 by Wilcoxon) and to purity adjustment (371/371 retained). LASSO-Cox yielded a 16-gene signature, although only 5 of the 16 genes were selected in ≥ 50% of CV folds (C6, PDE4D, CRTAC1, C4BPA and one lncRNA), so the specific gene membership is unstable. Cross-validated discrimination was modest but reproducible (out-of-fold [OOF] C-index 0.62, 95% CI 0.54–0.69; apparent C 0.73, 95% CI 0.67–0.79) and robust to the L1 penalty (OOF C 0.55–0.62 across λ; nested-CV λ selection 0.55). Using the honest cross-validated (out-of-fold, OOF) risk score, higher risk remained associated with OS after adjustment for stage and age (HR 1.34 per SD, 95% CI 1.06–1.68, p = 0.013), but the incremental discrimination over a stage + age model was very small (ΔC-index + 0.01); stage-independence is therefore statistical but marginal. When stratified by the OOF score the signal was carried by stage II–IV (HR 1.39, p = 0.015) and was not significant within stage I (HR 0.94, p = 0.79). Importantly, and in contrast to the optimistic in-sample result, the OOF risk score was attenuated to non-significance after adjustment for a 19-gene proliferation meta-score (HR 1.20, 95% CI 0.95–1.51, p = 0.13; with stage + age, HR 1.25, p = 0.08): the LOY prognostic signal is entangled with, and not cleanly separable from, generic proliferation. The signature reproduced in the male subset of GSE31210 (n = 105, 19 events; log-rank p = 0.048; 3-/5-year AUC 0.68/0.66), attenuated relative to the mixed-sex cohort as expected for a smaller, event-poor male subset. Conclusion. In male LUAD, LOY is a genomically-validated transcriptional signal that is statistically but only marginally independent of TNM stage (small incremental discrimination) and that is not cleanly separable from generic proliferation once honest cross-validated risk is used. The individual 16-gene model is only partly reproducible and should be regarded as one realisation of an LOY-anchored risk signal rather than a locked biomarker; prospective and larger male-specific validation, and clearer separation from proliferation, are required before clinical use.