Fig. 4: Autism traits and functional asymmetry.
From: Diverging asymmetry of intrinsic functional organization in autism
A Machine learning pipeline. We used 540 features (180 parcels * 3 gradients) to predict the total score of ADOS. All subjects were split into 4:1 training:test samples. Linear regression with elastic net (L1_ratio = 0.1) was the feature selector during which 5-fold cross validation was employed. We permuted this procedure 100 times by splitting the subjects randomly. B Shows how the model works in the training and testing samples using mean absolute error (MAE) during the 100 permutations. C Summarizes the frequency of this feature being selected across the 100 permutations. D Illustrates the distribution of correlations between the observed ADOS total score and predicted ADOS total score in the testing samples across the 100 permutations. Vis1: primary visual network, Vis2: secondary visual network, SMN somatomotor network, CON cingulo-opercular network, DAN dorsal attention network, Lan. language network, FPN frontoparietal network, Aud. auditory network, DMN default mode network, PMN posterior multimodal network, VMN ventral multimodal network, OAN orbito-affective network.
