Composition-Aware Analysis of Public Testis Transcriptomes Identifies Reproducible Tissue-State Signatures in Nonobstructive Azoospermia

Whole-testis transcriptomes can conflate loss of germ cells with transcriptional changes in retained somatic compartments. We used public human testis datasets to determine whether a composition-aware transcriptional pattern is reproducible in nonobstructive azoospermia (NOA). Donor-level single-cell analysis showed marked depletion of recovered spermatocyte and spermatid states in NOA. We held a germline signature and a GSE149512-derived stromal/macrophage-associated signature fixed during evaluation in two independent obstructive azoospermia (OA) comparator cohorts. In GSE145467 (10 OA and 10 NOA), NOA showed lower germline scores and higher tissue-state scores (Hedges’ g = 1.70 for the tissue-state score, 95% confidence interval [CI] 0.92–3.77; exact permutation p = 0.00108). The same direction was observed in GSE9210 (11 OA and 47 NOA; g = 3.69, 95% CI: 2.28–7.26; 100,000-draw permutation p = 0.000010). The original mixed-control bulk cohort was retained only as a within-cohort sensitivity analysis because its within-cohort remodeling axis was strongly correlated with the NOA label ( r = 0.899; variance inflation factor = 5.19). Small spermatocyte-focused data provided exploratory, rather than confirmatory, cell-state observations. These results support a reproducible bulk tissue-state pattern in NOA, but do not establish absolute cell-number changes, cell-intrinsic mechanisms, microdissection testicular sperm extraction prediction, or clinical utility.