The study of brain communication based on fMRI data is often limited because such measurements are a mixture of session-to-session variability with subject- and condition-related information. Disentangling these contributions is crucial for real-life applications, in particular when only a few recording sessions are available. The present study aims to define a reliable standard for the extraction of multiple signatures from fMRI data, while verifying that they do not mix information about the different modalities (e.g., subjects and conditions such as tasks performed by them). In particular, condition-specific signatures should not be contaminated by subjectrelated information, since they aim to generalize over subjects. Practically, signatures correspond to subnetworks of directed interactions between brain regions (typically 100 covering the whole brain) supporting the subject and condition identification for single fMRI sessions. The key for robust prediction is using effective connectivity instead of functional connectivity. Our method demonstrates excellent generalization capabilities for subject identification in two datasets, using only a few sessions per subject as reference. Using another dataset with resting state and movie viewing, we show that the two signatures related to subjects and tasks correspond to distinct subnetworks, which are thus topologically orthogonal. Our results set solid foundations for applications tailored to individual subjects, such as clinical diagnostic.

This work was supported by the European Union's Horizon 2020 research and innovation programme/Human Brain Project (grant FP7-FET-ICT-604102 to MG and GD; H2020-720270 HBP SGA1 to GD) and the Marie Sklodowska-Curie Action (grant H2020-MSCA-656547 to MG). A.I. is supported by the Spanish Ministry of Economy and Competitiveness Flag-ERA APCIN project CHAMPMouse (PCIN-2015-127). GD also acknowledges funding from the ERC Advanced Grant DYSTRUCTURE (#295129), the Spanish Research Project PSI2016-75688-P and the Catalan Research Group Support 2017 SGR 1545. DM was supported by the KU Leuven Special Research Fund (grant C16/15/070). SK has been funded by a Heisenberg grant from the German Science Foundation (DFG KU 3322/1-1), the European Union (ERC-2016-StG-Self-Control-677804) and a Fellowship from the Jacobs Foundation (JRF 2016-2018). This work has in part been funded by the German Science Foundation (SFB 936/C7).