Multicohort transcriptomic and machine-learning analysis identified 53 shared candidate genes between NAFLD and MI, five of which were retained by all three feature-selection methods, and exploratory qPCR showed IDI1 expression was higher in MI patients and in mouse blood and infarcted myocardium, supporting an association between IDI1 and myocardial injury.
Key Findings
Methods
The study integrated six GEO microarray cohorts spanning NAFLD and MI to look for shared transcriptomic features.
Six microarray cohorts from the GEO database were analyzed.
Candidate genes were defined using weighted gene co-expression network analysis (WGCNA), differential expression analysis, and fuzzy c-means clustering across ordered cross-sectional NAFLD groups.
The aim was to identify shared disease-associated transcriptomic features across independent NAFLD and MI populations.
Results
Fifty-three shared candidate genes were identified between NAFLD and MI.
The candidates came from the combined WGCNA, differential expression, and fuzzy c-means clustering approaches.
These genes were then passed to machine-learning feature selection.
Results
Five genes were retained by all three machine-learning feature-selection methods.
The three methods were LASSO regression, random forest, and SVM-RFE.
The overlapping genes were treated as consensus computational candidates.
Candidates were examined with exploratory classification, ROC curves, calibration curves, decision curve analysis, and SHapley Additive exPlanations (SHAP).
Results
Single-gene ROC analyses highlighted SULF2, GADD45A, LIN7A, and IDI1 within the analyzed case-control datasets.
The abstract reports these four genes as showing discriminative performance in single-gene ROC analyses.
Analyses were limited to the case-control datasets analyzed, and the classification work was described as exploratory.
Results
IDI1 expression was higher in patients with established MI than in healthy controls.
qPCR was used for exploratory expression assessment in patients with established MI.
The difference was reported as p = 0.004.
Results
IDI1 expression was elevated in blood and infarcted myocardium of mice after LAD ligation.
A mouse model of acute ischemic myocardial injury was created by LAD ligation.
Blood IDI1 expression was higher, p = 0.034.
Infarcted myocardium IDI1 expression was higher, p = 0.0072.
Conclusions
The authors conclude that the findings indicate shared transcriptomic links between NAFLD and MI and support an association between IDI1 and myocardial injury.
IDI1 was selected for exploratory expression assessment rather than as a validated biomarker.
The experimental validation was framed as exploratory, in both human MI patients and an acute mouse injury model.
What This Means
This study asked whether non-alcoholic fatty liver disease (NAFLD) and heart attack (myocardial infarction, MI) share common gene-activity patterns. The researchers combined six publicly available gene-expression datasets and used several statistical and machine-learning methods to narrow down the genes involved. They found 53 genes linked to both conditions, and five of these were picked by all three machine-learning selection methods used.
One of these genes, IDI1, was tested in the laboratory. Its activity was higher in blood from patients who had had a heart attack compared with healthy people. In mice with a surgically induced heart attack, IDI1 was also higher in both blood and the damaged heart tissue. Three other genes (SULF2, GADD45A, LIN7A) also showed an ability to distinguish cases from controls in the datasets analyzed.
This research suggests there may be shared molecular links between fatty liver disease and heart attack, and that IDI1 may be associated with heart injury. The authors describe the experimental work as exploratory, so the findings do not show that IDI1 causes heart injury or that it is ready to be used as a diagnostic test; further studies would be needed.
Qiao J, Chen S, Li S, Shao W, Zhang Y, Huang Y, et al.. (2026). Cross-Disease Transcriptomic Signatures Shared by NAFLD and MI: Multicohort Machine-Learning Analysis and Exploratory IDI1 Expression Assessment.. International journal of molecular sciences. https://doi.org/10.3390/ijms27209390