Tracing high transductive cohort AUC to same-site supervision in a site-aware population GNN for multisite fMRI
- 1National Center for Applied Mathematics, Chongqing Normal University
Abstract
Reliable population-graph evaluation requires separating cohort-conditioned discrimination from performance at an unseen site. We evaluate a recovered site-aware graph neural network in ABIDE-I (871 participants, 20 sites), distinguishing archived intervention results from newly recomputed controls. Archived cohort AUC was approximately 0.94, with high discrimination retained by site-only connectivity. A reconstructed ten-fold empirical site-prior classifier achieved pooled out-of-fold AUC 0.506 (95\% site-resampling interval [0.418, 0.546]); simple site proportions therefore did not reproduce the high cohort result. In a budget-matched support-location experiment, the across-site mean AUC difference was +0.252 [0.184, 0.322] with directed edges but -0.003 [-0.020, 0.015] after symmetrisation, demonstrating implementation sensitivity rather than a stable support-location advantage. A supplementary leave-one-site-out analysis compared scoring-time population-edge access after the same source-site training procedure. Mean site-macro AUC was 0.577 with access and 0.580 without it; the paired difference was -0.0037 [-0.0179, 0.0111]. These observations do not establish equivalence or a complete mechanism. Explicit access contracts, compatible estimators and run-level provenance are necessary to interpret high transductive cohort performance.
Related
- The contribution of population genetics to the epidemiological tracing of pathogenic bacteria: the predominant clonal evolution model
- FODMAP Consumption by Adults from the French Population-Based NutriNet-Santé Cohort
- The influence of breastfeeding and pacifier use on the association between preterm birth and primary-dentition malocclusion: A population-based birth cohort study
- A cohort-based population model for the Poultry Red Mite Dermanyssus gallinae (Acari: Dermanyssidae): Application to explore potential pathways for control strategies using Fluralaner as a case study
- The Association Between APOE Genotype, Race, and Dementia: An Analysis of 7 Population-Based Cohort Studies
BibTeX
@misc{wan2026tracing,
title = {{Tracing high transductive cohort AUC to same-site supervision in a site-aware population GNN for multisite fMRI}},
author = {K. Wan and Z. Chen and G. Liu and B. Yu and Q. Zhang and F. Zhang and N. Zhong and H. Kuai},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.15.751719},
url = {https://www.biorxiv.org/content/10.64898/2026.09.15.751719v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Tracing high transductive cohort AUC to same-site supervision in a site-aware population GNN for multisite fMRI
AU - K. Wan
AU - Z. Chen
AU - G. Liu
AU - B. Yu
AU - Q. Zhang
AU - F. Zhang
AU - N. Zhong
AU - H. Kuai
PY - 2026
DA - 2026/09/15
PB - bioRxiv
DO - 10.64898/2026.09.15.751719
UR - https://www.biorxiv.org/content/10.64898/2026.09.15.751719v1
AB - Reliable population-graph evaluation requires separating cohort-conditioned discrimination from performance at an unseen site. We evaluate a recovered site-aware graph neural network in ABIDE-I (871 participants, 20 sites), distinguishing archived intervention results from newly recomputed controls. Archived cohort AUC was approximately 0.94, with high discrimination retained by site-only connectivity. A reconstructed ten-fold empirical site-prior classifier achieved pooled out-of-fold AUC 0.506 (95\% site-resampling interval [0.418, 0.546]); simple site proportions therefore did not reproduce the high cohort result. In a budget-matched support-location experiment, the across-site mean AUC difference was +0.252 [0.184, 0.322] with directed edges but -0.003 [-0.020, 0.015] after symmetrisation, demonstrating implementation sensitivity rather than a stable support-location advantage. A supplementary leave-one-site-out analysis compared scoring-time population-edge access after the same source-site training procedure. Mean site-macro AUC was 0.577 with access and 0.580 without it; the paired difference was -0.0037 [-0.0179, 0.0111]. These observations do not establish equivalence or a complete mechanism. Explicit access contracts, compatible estimators and run-level provenance are necessary to interpret high transductive cohort performance.
ER -