Moderate evidence
More
AI tells unstable from stable heart disease in gene data, but only in computer tests
Study title: Integrative Single-Cell Transcriptomics, Multi-Omics Analyses, and Computational Pharmacology Reveal Macrophage Heterogeneity and the Putative THBS1-CD36 Axis in Coronary Atherosclerosis
- Evidence type
- Controlled or validated comparison
- Sources
- 1 sourcePeer-reviewed
- Published
- 1 min read
- Truvace record
- TRV-2026-1245cite or check it
Study title: Integrative Single-Cell Transcriptomics, Multi-Omics Analyses, and Computational Pharmacology Reveal Macrophage Heterogeneity and the Putative THBS1-CD36 Axis in Coronary Atherosclerosis
Researchers constructed a single-cell atlas of human coronary atherosclerosis and identified six macrophage subpopulations, focusing on a THBS1+ subset linked to hypoxia adaptation and TGF-β signaling. They derived a 12-module co-expression network with NEW10 as a THBS1+ signature, predicted THBS1-CD36 crosstalk with lymphatic endothelial cells, and computationally screened SMS121 against CD36.
Read the study abstract
Coronary atherosclerosis (CA) is characterized by profound macrophage heterogeneity that drives plaque progression and vulnerability, yet the precise subpopulations and their niche-specific functions remain incompletely defined. Here, we constructed a single-cell transcriptomic atlas of human CA using dataset GSE131778 and identified six distinct macrophage subpopulations. Among these, the THBS1+ macrophage subset emerged as a terminally differentiated, hypoxia-adaptive population with elevated TGF-β signaling a…
The gain
Machine-learning models built on a macrophage gene signature separated acute coronary syndrome from stable disease and ruptured from stable plaques, reaching AUC 0.9333 in an independent cohort, in computer analysis only.
In brief
By October 2026, the team had trained 13 machine learning classifiers on NEW10 genes to distinguish acute coronary syndrome from stable disease and ruptured from stable plaques, reporting AUC 0.9333 for SVM linear and Naive Bayes in an independent cohort. The work remains in silico, positioning the signature and predicted THBS1-CD36 axis as candidates for future mechanistic and pharmacological validation rather than deployed clinical tools.
Main points
- Single-cell atlas from dataset GSE131778 identified six distinct macrophage subpopulations in human coronary atherosclerosis.
- THBS1+ macrophages were described as terminally differentiated, hypoxia-adaptive with elevated TGF-β signaling and FOSB-driven regulation.
- High-dimensional WGCNA identified 12 modules, with module NEW10 as specific signature of THBS1+ macrophages.
- Cell-cell communication predicted THBS1-CD36 ligand-receptor crosstalk between THBS1+ macrophages and lymphatic endothelial cells.
- Molecular docking and 100-ns molecular dynamics predicted favorable stable interaction between SMS121 and CD36.
The full story
The study built a single-cell transcriptomic atlas of human coronary atherosclerosis from GSE131778, defining six macrophage subsets and singling out THBS1+ macrophages as hypoxia-adaptive with TGF-β signaling. WGCNA produced 12 transcriptional modules, with NEW10 marking THBS1+ cells.
Communication analysis predicted THBS1-CD36 interaction with lymphatic endothelial cells, while docking and 100-ns MD simulations suggested SMS121 binds CD36 stably. The NEW10 signature was then used to train classifiers for ACS versus sCAD and ruptured versus stable plaque discrimination.
Sources
1 on file- Chemical Biology & Drug Designdoi:10.1111/cbdd.70399
Talk it over
The classifiers reached AUC 0.9333 in an independent cohort, but only in computer analysis of gene data. If you were a cardiologist picking between patients with acute coronary syndrome and stable disease, what would you need to see before this number meant anything for them?
AI tells unstable from stable heart disease in gene data, but only in computer testshttps://truvace.com/item/integrative-single-cell-transcriptomics-multi-omics-analyses-and-computational-p
The classifiers reached AUC 0.9333 in an independent cohort, but only in computer analysis of gene data. If you were a cardiologist picking between patients with acute coronary syndrome and stable disease, what would you need to see before this number meant anything for them?
Answer in the threadThe atlas came from one dataset, GSE131778. Who or what could be missing from that data, and how might that change how well the NEW10 signature separates ruptured from stable plaques?
Answer in the threadResearchers call the THBS1-CD36 axis and SMS121 candidates for future validation. Which step should come next, and who should decide whether it is worth the lab time?
Answer in the thread
If you have more time
The team trained 13 machine learning classifiers and reported AUC 0.9333 for SVM linear and Naive Bayes. What would you want to know about how those were picked out before trusting the number?
Answer in the threadWhat kind of experiment or patient study would settle whether THBS1+ macrophages actually drive unstable heart disease, rather than just marking it in the data?
Answer in the thread
ace
What readers say