Network and State Specificity in Connectivity-Based Predictions of Individual Behavior

Abstract

Predicting individual behavior from brain functional connectivity (FC) patterns can contribute to our understanding of human brain functioning. This may apply in particular if predictions are based on features derived from circumscribed, a priori defined functional networks, which improves interpretability. Furthermore, some evidence suggests that task-based FC data may yield more successful predictions of behavior than resting-state FC data. Here, we comprehensively examined to what extent the correspondence of functional network priors and task states with behavioral target domains influences the predictability of individual performance in cognitive, social, and affective tasks. To this end, we used data from the Human Connectome Project for large-scale out-of-sample predictions of individual abilities in working memory (WM), theory-of-mind cognition (SOCIAL), and emotion processing (EMO) from FC of corresponding and non-corresponding states (WM/SOCIAL/EMO/resting-state) and networks (WM/SOCIAL/EMO/whole-brain connectome). Using root mean squared error and coefficient of determination to evaluate model fit revealed that predictive performance was rather poor overall. Predictions from whole-brain FC were slightly better than those from FC in task-specific networks, and a slight benefit of predictions based on FC from task versus resting state was observed for performance in the WM domain. Beyond that, we did not find any significant effects of a correspondence of network, task state, and performance domains. Together, these results suggest that multivariate FC patterns during both task and resting states contain rather little information on individual performance levels, calling for a reconsideration of how the brain mediates individual differences in mental abilities.

Keywords: brain-based prediction; brain–behavior relationships; fMRI; functional connectivity; interindividual differences; machine learning.

Figure 1.

Overview of the sampling procedure.

Figure 2.

Overview of the applied methods. Yellow blocks depict the network extraction from sample 1. Violet blocks depict the network based prediction in sample 2, together with the feature extraction (functional connectivities) from the networks delineated in the first step and sample. The upper heatmap under “FC-Features” shows the FC from the different states in the WM-network. the GLM: General linear modelling, PE: Phase encoding, FC: Functional connectivity, Soc. Cog.: Social cognition task (in-scanner task), Soc. Satisf.: Social Satisfaction Questionnaire (out-of-scanner score), Emo. Match.: Emotional Face Matching task (in-scanner task), Emo. Recog.: Emotional Face Recognition task (out-of-scanner score).

Figure 3.

Prediction performance: Boxplots of the distribution of COD and RMSE from the 100 x CV for each phenotypic domain (A – WM, B – Social, C – EMO domain), state and network. Bars represent the model fit / COD (all negative values are set to 0) and RMSE of prediction of a specific score (WM, SOCIAL, EMO; performed in (darker background) or outside (lighter background) the scanner) based on functional connectivity within a given network (POWER, WM, SOCIAL, EMO) in a given state (REST, WM, SOCIAL, EMO). Green: WM, blue: SOCIAL, red: EMOTION, yellow: resting state, white: Power nodes. Darker background: target is the task performed in the scanner; lighter background: target is the task performed outside the scanner. Soc. Cog.: Social cognition task (in-scanner task), Soc. Satisf.: Social Satisfaction Questionnaire (out-of-scanner score), Matching: Emotional Face Matching task (in-scanner task), Recognition: Emotional Face Recognition task (out-of-scanner score).

Figure 4.

State and network specificity: State (A-C) and network (D-E) specificity for each phenotypic domain (A and D – prediction of WM scores, B and E – prediction of SOCIAL scores, C and F – prediction of EMO scores). For state specificity (A-C) RMSE of all networks (POWER, WM, SOCIAL, EMO) and the two tasks of the respective domain in a given state (REST, WM, SOCIAL, EMO) are averaged. For network specificity all states (REST, WM, SOCIAL, EMO) and the two task of the respective domain are averaged in a given network (Power nodes, WM, SOCIAL, EMO). Green: WM, blue: SOCIAL, red: EMOTION, yellow: resting state. Horizontal bars indicate significance pcorr < 0.05.

Figure 5.

State–target similarity: Boxplots of the distribution of RMSE RMSE from the 100 x CV averaged across all networks (Power nodes, WM, SOCIAL, EMO network) within a given state (WM, SOCIAL and EMO) and task (SAME or SIMILAR) Green: WM, blue: SOCIAL, red: EMOTION, grey: averaged across networks. Darker background: target is the “same” task performed in the scanner, lighter background: target is the “similar” task performed outside the scanner.