AI Science Tracker
EN

Leukocyte RNA retention modestly changes platelet-based cancer classification

University of Gdansk; Medical University of Gdansk; Gdansk University of Technology; Polish Academy of Sciences; University of Utah; Amsterdam UMC · 30 Oct 2026 · medicine

Discovery

Leukocyte RNA retention modestly changes platelet-based cancer classification

Source check: Source checked. Not independent scientific confirmation.

Finding: Ten learned classifiers tested whether leukocyte-associated RNA retention affects platelet RNA cancer detection. In a matched 754-sample analysis, training on high-retention samples modestly improved some classifiers; most differences were not significant. Adding the retention score to a full-cohort classifier changed AUC only from 0.9247 to 0.9249, challenging a blanket assumption that this signal should always be removed.

Technology: SVM; LightGBM; XGBoost; random forest and other learned classifiers

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned classifiers provided the cancer-detection experiments needed to quantify how differing RNA-retention strata change diagnostic performance; the finding concerns assay signal composition rather than a new general classifier architecture.
Test and result
Analysis combined 2,351 public platelet RNA samples with a marker panel derived from single-cell data from 29 donors. Matching yielded 377 high-retention and 377 low-retention samples, with cancer type balanced but residual collection-center imbalance. Ten classifiers used separate training/validation/test sets across repeated splits and validation-only hyperparameter optimization. SVM mean test AUC was 0.76 when trained on low-retention versus 0.84 on high-retention samples; corresponding LightGBM values were 0.74/0.79 and XGBoost 0.71/0.74. Most other training-retention comparisons were not significant.
Limitation / still unknown
RNA retention score is a compositional proxy, not physical purity or a causal manipulation. Residual collection-center confounding remains; repeated overlapping splits do not provide 30 independent biological replications. Model-specific modest effects do not establish universally beneficial retention, a correction standard or clinical benefit. The full-cohort added-score comparison was nearly null, and generalization to separately collected clinical cohorts is untested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13722837/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 21 Nov 2026. Review scope: Read actual primary marker/data and matched-cohort definitions, ten named classifiers, splitting/tuning Methods, actual performance tables 1–2, Results and interpretive limitations. Operational sample-processing procedures and supplements are outside the represented evidence scope. Exact primary identity/date and fresh primary PubMed correction/retraction relationships checked..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse medicine discoveries.