Skip to main content
. 2022 Jun 3;12:228. doi: 10.1038/s41398-022-01987-x

Fig. 3. ROC curve for prediction for the immuno-behaviourally defined responding group.

Fig. 3

The classifiers included the Oblique Random Forest (ORF) model, Partial Least Squares (PLS) model, Support Vector Machine (SVM) model, sparse Linear Discriminant Analysis (sLDA) model and Neural Networks (NN) model. Based on the immuno-behavioural covariation plane, the models were trained to predict the response to bumetanide for the children with ASD. As described in the main text, the models were trained using the Discovery Set, and tested using the Validation Set. The performances of the classification accuracy in the testing data set were reported in this figure. A Models with the cytokine levels at the baseline for predicting patients with ASD in the best responding group. B Models without the cytokine levels at the baseline for predicting patients with ASD in the best responding group. C Models with the cytokine levels at the baseline for predicting patients with ASD in the least responding group. D Models without the cytokine levels at the baseline for predicting patients with ASD in the least responding group.