Re-analysis of five public transcriptomic datasets identified overlapping exploratory candidate signals between coronary artery disease and non-obstructive azoospermia cohorts, but did not establish a shared causal pathway, clinical diagnostic utility, or direct correspondence between testicular and coronary cell states.
Key Findings
Results
Nominal differential-expression screening identified 978 CAD-associated and 2562 NOA-associated candidate transcripts across separately analysed public transcriptomic cohorts.
Five Gene Expression Omnibus (GEO) datasets were re-analysed
Differential-expression screening was used as a nominal screening tool
978 transcripts were identified as CAD-associated candidates
2562 transcripts were identified as NOA-associated candidates
The analyses were conducted on separate CAD and NOA cohorts without assuming a shared causal mechanism
Results
WGCNA identified a moderate correlation between the MElightyellow module and NOA status in dataset GSE45887.
The correlation between the MElightyellow module and NOA status was r = 0.58, p = 0.007
GSE45887 was described as 'a small, imbalanced, non-independent subset of GSE45885'
The authors note this finding comes from a limited and methodologically constrained dataset
Weighted gene co-expression network analysis (WGCNA) was the analytical method used
Results
An archived neural-network ranking prioritised five candidate genes — HSPA1B, PLCL2, ISLR2, STRN, and AQP7 — for descriptive analyses.
The neural-network (NNET) ranking was generated from the same 20 specimens used in other analyses
The authors explicitly treat this ranking as 'heuristic' rather than confirmatory
These five genes were prioritised for descriptive purposes only
The archived nature of the ranking means it was not independently validated in a separate cohort
Results
xCell enrichment analysis produced marker-gene enrichment scores rather than direct measurements of cell abundance or function.
xCell was applied to the transcriptomic datasets for cell-type enrichment scoring
The authors clarify that xCell scores represent marker-gene enrichment, not direct cell quantification
This methodological limitation constrains interpretation of any apparent differences in cell-type composition between groups
Results
Single-cell transcriptomic mapping using GSE149512 was limited to descriptive conclusions due to heterogeneous sample composition.
GSE149512 combined heterogeneous NOA aetiologies with paediatric and adult comparator tissues
The mixture of aetiologies and age groups precluded definitive cell-state comparisons
Mapping was explicitly described as 'descriptive' in scope
No direct correspondence between testicular and coronary cell states could be established
Conclusions
The study explicitly does not establish a shared causal pathway, sex- or age-independent association, temporal sequence, clinical diagnostic utility, or direct correspondence between testicular and coronary cell states.
Authors state: 'The analyses generate hypotheses from separate CAD and NOA cohorts'
Authors explicitly state the analyses 'do not establish a shared causal pathway'
No claims of clinical diagnostic utility are made
The study is framed as exploratory and hypothesis-generating rather than confirmatory
What This Means
This research suggests that when large public gene expression datasets from patients with coronary artery disease (CAD, a condition affecting heart arteries) and non-obstructive azoospermia (NOA, a cause of male infertility where sperm are not produced) are analysed separately and then compared, some overlapping gene signals can be identified. The researchers used several computational methods — including network analysis, a neural-network ranking tool, cell-type scoring, and single-cell mapping — to explore whether any transcriptomic (gene activity) patterns appeared in both conditions. They identified nearly 1,000 CAD-linked and over 2,500 NOA-linked candidate gene transcripts, and found a moderate statistical association between one gene network module and NOA status, as well as five prioritised candidate genes.
However, the researchers are careful to frame all of these findings as preliminary and exploratory. Each analysis method came with notable limitations: the datasets were small, some were subsets of each other, the neural-network ranking was generated from only 20 samples, and the single-cell dataset mixed different types of NOA with tissues from both children and adults. The study does not claim that CAD and NOA share a common biological cause, nor does it suggest these gene patterns could be used to diagnose either condition.
This research matters because it illustrates both the potential and the pitfalls of mining publicly available genomic data to look for unexpected connections between seemingly unrelated diseases. The findings are best understood as a starting point for generating new research questions — for example, whether specific genes like HSPA1B or AQP7 might play roles in both heart and reproductive tissue biology — rather than as evidence of a proven biological link between CAD and NOA.
Wang T, Song P, Wang H, Wang H, Zou R, Guo S. (2026). Exploratory Cross-Cohort Transcriptomic Comparison of Coronary Artery Disease and Non-Obstructive Azoospermia.. International journal of molecular sciences. https://doi.org/10.3390/ijms27177909