The abstract is read at the publisher; the summary is JClub's.
XGBoost trained on TVAE-augmented circulating miRNA data for glioblastoma achieved AUROC 0.59 (95% CI 0.49–0.70) on GSE184472 (n=118), but performance did not generalize to CGGA.
- Why it matters: Small discovery cohorts make it hard to separate true biological signal from cohort- or platform-specific artifacts, so same-modality validation can overstate diagnostic value.
- What they did: CTGAN and TVAE were trained on GSE122488 (38 patients) to generate 500 synthetic samples each, then Random Forest and XGBoost were evaluated using escalating validation across GSE184472 (n=118) and CGGA tissue data (n=198).
- The result: AUROC on CGGA was 0.40 (95% CI 0.32–0.48) before Min-Max harmonisation and 0.47 after; after harmonisation, recall reached 1.00 by predicting all samples as GBM, yielding accuracy 0.41.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.