Fig. 6: Performance of the LASSO model used to construct a miRNA signature to diagnose sCJD. | Communications Biology

Fig. 6: Performance of the LASSO model used to construct a miRNA signature to diagnose sCJD.

From: Distribution of microRNA profiles in pre-clinical and clinical forms of murine and human prion disease

Fig. 6

A LASSO coefficient profiles of the 14 miRNAs including age and a random variable generated using the training cohort. Each curve corresponds to a miRNA. The vertical line is drawn at the value λ = 0.0794 chosen by 1000x cross-validation. B Receiver operating characteristic (ROC) curve for the performance of miRNAs and AUC values calculated using the validation set. C Correlation matrix heatmap of the 14 miRNAs from the training cohort, where each cell represents the Pearson correlation between the miRNA indicated in the corresponding row and columns. Dendrograms show the hierarchical clustering of the collinearity of the 14 candidate miRNAs. D The variable inclusion plots that indicate the three most important features above reference variable (RV). E–G Model stability bubble plots representing the stability of the model with the presence and absence of the strongest miRNAs in the model, hsa-miR-423-3p, hsa-miR-101-3p and hsa-miR-1306-5p. Each bubble represents a model and the colour of the bubble indicates whether (red) or not (blue) the model includes the particular miRNA. The size of the bubble corresponds to the probability of the model of the same size obtained from the ×1000 cross-validation.

Back to article page