Answers are grounded in this paper and cite their sources.
Reuse of data-independent acquisition (DIA) proteomics can increase analytical reach, but inconsistent sample identities, dependent cohorts, small numbers of independent studies, and optimistic machine-learning evaluation can undermine statistical validity. We developed a source-aware framework that links biological-unit provenance to parent-resource meta-analysis and cohortheld-out prediction. DIA-NN-derived quantitative matrices and source annotations from PXD016278, PXD043216, PXD035249, and PXD063058 were analyzed as four independent cerebrospinal-fluid (CSF) parent resources. PXD067114 provided a separate temporal-lobe brain comparison. The framework used source-scoped specimen and participant sample records, prespecified coverage rules, Hedges' g, two-stage restricted maximum likelihood (REML), modified Knapp-Hartung (mKH) inference, completefamily Benjamini-Hochberg adjustment, numerical verification, simulations, biological sensitivities, and leakage-controlled cohortheld-out machine learning. The analysis retained 598 CSF participants across six cohorts and 114 brain specimens. Of 663 common proteins, 637 were estimable across all four CSF parents. Exploratory REML-normal analysis identified 29 BH associations, whereas primary mKH inference yielded 0 BH discoveries. The smallest primary p-value was 0.00536 for SMOC1 and the smallest primary q-value was 0.882. Six cohort-held-out rank Elastic-Net AUCs ranged from 0.754 to 0.952; AUC was 0.779 for MGH and 0.769 for PXD063058. Prediction replay, held-out-label intervention, and 30 complete training-label permutation refits passed their prespecified checks. The results support the framework as a reproducible method for combining DIA-NN-derived studies while keeping confirmatory inference, exploratory biological evidence, and cohort-held-out prediction clearly separated.
No takes yet. Share an insight, caveat, or question.
Karanikolaos et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: