Fig. 4

Robustness and replicability summary. a The histogram summarizes the percentage of clusters showing a robust and replicable pattern of results as that observed in discovery analyses in the ABIDE sample for main effects of diagnosis (Dx; green; N = 7 clusters), sex (yellow; N = 10 clusters) and their interaction (blue; N = 1 cluster) across three R-fMRI metrics. All findings were robust to different preprocessing pipelines. Across R-fMRI metrics, main sex effects were moderately (50%) to largely (80%) replicable across independent samples: Gender Explorations of Neurogenetics and Development to Advance Autism Research (GENDAAR) and the EU-AIMS Longitudinal European Autism Project (LEAP), respectively. Replicability for main effects of diagnosis was largely replicable in GENDAAR (86%) and minimally replicable in EU-AIMS LEAP (29%). The VMHC pattern observed for sex-by-diagnosis interaction in discovery analyses was replicated in EU-AIMS LEAP only. b Surface conjunction maps show the clusters replicated in GENDAAR only (G, purple), EU-AIMS LEAP only (E, blue) and in both samples (G and E, red) for each effect separately