Cross study transfer of vascular transcriptomic classifiers in seven public cohorts

Source-rank L1 logistic regression discriminated vascular transcriptomic classes with AUC 0.947 in nested validation, but external AUCs dropped to 0.753 and 0.667 across carotid cohorts.

• Why it matters: Strong within-study discrimination may fail when clinical endpoints, tissue compartments, or assay preprocessing differ, limiting whether classifiers generalize across vascular contexts.
• What they did: The work tested four conventional classifiers plus a prior-only control across seven GEO datasets (294 observations) using nested validation in 48 abdominal aortic aneurysm samples, then external evaluation under three preprocessing configurations.
• The result: External performance was 1.000 in 11 macrophage plaque samples but fell to 0.367 under a separately source-tuned common-gene pipeline, and survival prediction was weak (Harrell C 0.565, 0.524 after 30-day landmark).

AI summary of the abstract · may contain errors · not medical advice · read the paper.