The study team combined public bulk transcriptomic datasets with a curated list of adenosine-related genes to search for molecular contributors to coronary heart disease (CHD). They first performed differential expression analysis and identified 4,884 differentially expressed genes. A weighted gene co-expression network analysis then isolated 243 ARG-associated module genes, producing 231 candidate genes taken forward for feature selection. Investigators applied multiple machine-learning filters—XGBoost, support vector machine-recursive feature elimination, and Boruta—to narrow the list. After these steps, DYRK2 and SIGIRR emerged as prioritized genes that were consistently downregulated in CHD bulk datasets, including a separate validation cohort mentioned in the preprint abstract.
The authors reported a focused, data-driven pipeline. The abstract does not present detailed parameter settings for the network analysis or the training/validation splits used by the machine-learning models. Those methodological specifics were not reported in the abstract and therefore are not reproduced here.
From the prioritized genes, the investigators constructed a two-gene nomogram using DYRK2 and SIGIRR as diagnostic features. The preprint states the model demonstrated strong diagnostic performance across datasets. The study team reported consistent downregulation of both genes in discovery and validation bulk cohorts. The abstract does not include numerical performance metrics, confidence intervals, or independent prospective validation data; those details were not provided in the source material.
The diagnostic approach, as described, leveraged cross-dataset consistency and machine-learning selection to support the two-gene panel. The preprint did not report how the nomogram would perform in routine clinical samples or in broader, demographically diverse cohorts.
Functional enrichment and pathway analyses connected DYRK2 and SIGIRR to processes implicating mitochondrial dysfunction and innate immune regulation. Data indicated a shift toward immune–metabolic imbalance within CHD tissue. Immune infiltration analyses described reduced adaptive immune cell populations alongside enhanced myeloid-associated inflammatory signals in CHD samples. The study team observed these systemic trends across the integrated transcriptomic datasets.
Authors noted these associations rather than proving causal relationships. The abstract does not include mechanistic perturbation experiments or direct assays of mitochondrial function tied to the two genes. Those experimental confirmations were not reported in the abstract.
Single-cell RNA sequencing localized altered expression of DYRK2 and SIGIRR predominantly to the vascular smooth muscle cell compartment. The single-cell analysis traced expression changes along pseudotime trajectories, suggesting dynamic remodeling within smooth muscle populations during CHD-associated phenotypic switching. Investigators reported concurrent remodeling of vascular endothelial growth factor and pleiotrophin signaling within these cellular trajectories.
These single-cell data link adenosine-related gene expression to cellular states that are central to vascular remodeling. The preprint abstract does not specify the number of single-cell libraries, sequencing depth, or the precise pseudotime inference method; those methodological points were not provided in the abstract and are therefore not restated here.
To validate the bioinformatic findings, the authors performed reverse transcription-quantitative PCR on clinical samples. The abstract reports that RT-qPCR in clinical samples (n = 5 per group) supported the direction of the expression changes observed in silico, with P < 0.05. This small, targeted validation corroborated the downregulation of DYRK2 and SIGIRR in CHD tissue as detected in bulk and single-cell analyses.
Because the validation cohort was limited to five samples per group, the RT-qPCR results should be interpreted as preliminary support rather than definitive confirmation across diverse patient populations.
This work is presented as a preprint on bioRxiv and has not been peer-reviewed. The authors declared no competing interests. Key limitations that the source explicitly reports or implies in the abstract include: the absence of detailed methodological parameters for some computational steps in the abstract, a small RT-qPCR validation sample size (n = 5 per group), and the lack of published numerical performance metrics for the two-gene nomogram within the abstract. The abstract also does not report in vitro or in vivo functional experiments that directly test causal roles for DYRK2 or SIGIRR in smooth muscle phenotypic switching.
These results identify DYRK2 and SIGIRR as candidate adenosine-related genes associated with immune–metabolic imbalance and smooth muscle cell dysfunction in CHD, according to the datasets and analytic workflow summarized in the preprint. Further mechanistic work and larger, peer-reviewed validation studies are necessary to establish clinical utility and biological causality.