No abstract

The prediction of inter-individual behavioural differences from neuroimaging data is a rapidly evolving field of research focusing on individualised methods to describe human brain organisation on the single-subject level. One method that harnesses such individual signatures is functional connectome fingerprinting, which can reliably identify individuals from large study populations. However, the precise relationship between functional signatures underlying fingerprinting and behavioural prediction remains unclear. Expanding on previous reports, here we systematically investigate the link between discrimination and prediction on different levels of brain network organisation (individual connections, network interactions, topographical organisation, and connection variability). Our analysis revealed a substantial divergence between discriminatory and predictive connectivity signatures on all levels of network organisation. Across different brain parcellations, thresholds, and prediction algorithms, we find discriminatory connections in higher-order multimodal association cortices, while neural correlates of behaviour display more variable distributions. Furthermore, we find the standard deviation of connections between participants to be significantly higher in fingerprinting than in prediction, making inter-individual connection variability a possible separating marker. These results demonstrate that participant identification and behavioural prediction involve highly distinct functional systems of the human connectome. The present study thus calls into question the direct functional relevance of connectome fingerprints.

Current major efforts in human neuroimaging research aim to understand individual differences and identify biomarkers for clinical applications. One particularly promising approach is the prediction of individual-level behavioural phenotypes (e.g. treatment response, cognition) from brain imaging data. An essential prerequisite to identify replicable brain-behaviour prediction models is sufficient measurement reliability. By attenuating the relationship between two variables, low reliability increases the sample size necessary to identify an effect, making large datasets a necessity rather than an advantage. While previous work has evaluated the reliability of brain-based measures, the impact of the reliability of behavioural phenotypes has been largely neglected, as target selection for prediction is often guided by scientific interest or data availability. Here we demonstrate the impact of low phenotypic reliability on individual-level prediction performance. Using simulated and empirical data from the Human Connectome Projects, we found that even moderate reliability levels of commonly used behavioural phenotypes can markedly limit the ability to link brain and behaviour when underlying relations are weak. Next, using 5000 subjects from the UK Biobank, we show that highly reliable data in smaller samples outperform large amounts of moderately reliable data. These findings suggest that research programmes focused on identifying generalizable brain-behaviour associations must seriously consider the reliability of outcome measures. Ultimately, a stronger focus on reliability will help reduce the financial and societal costs incurred in acquiring large-scale datasets with unreliable "markers" of behaviour.

The human brain operates in large-scale functional networks, collectively subsumed as the functional connectome1-13. Recent work has begun to unravel the organization of the connectome, including the temporal dynamics of brain states14-20, the trade-off between segregation and integration9,15,21-23, and a functional hierarchy from lower-order unimodal to higher-order transmodal processing systems24-27. However, it remains unknown how these network properties are embedded in the brain and if they emerge from a common neural foundation. Here we apply time-resolved estimation of brain signal complexity to uncover a unifying principle of brain organization, linking the connectome to neural variability6,28-31. Using functional magnetic resonance imaging (fMRI), we show that neural activity is marked by spontaneous "complexity drops" that reflect episodes of increased pattern regularity in the brain, and that functional connections among brain regions are an expression of their simultaneous engagement in such episodes. Moreover, these complexity drops ubiquitously propagate along cortical hierarchies, suggesting that the brain intrinsically reiterates its own functional architecture. Globally, neural activity clusters into temporal complexity states that dynamically shape the coupling strength and configuration of the connectome, implementing a continuous re-negotiation between cost-efficient segregation and communication-enhancing integration9,15,21,23. Furthermore, complexity states resolve the recently discovered association between anatomical and functional network hierarchies comprehensively25-27,32. Finally, brain signal complexity is highly sensitive to age and reflects inter-individual differences in cognition and motor function. In sum, we identify a spatiotemporal complexity architecture of neural activity — a functional "complexome" that gives rise to the network organization of the human brain.