Fig. 4: ML model for sPD/LRRK2-PD differentiation based on the transcriptomic profiles of the LRRK2int. | npj Parkinson's Disease

Fig. 4: ML model for sPD/LRRK2-PD differentiation based on the transcriptomic profiles of the LRRK2int.

From: Transcriptomics and weighted protein network analyses of the LRRK2 protein interactome reveal distinct molecular signatures for sporadic and LRRK2 Parkinson’s Disease

Fig. 4

A The logistic regression model with LASSO (Least Absolute Shrinkage and Selection Operator) was adopted to reduce dimensionality and select the most significant expression profiles for the LRRK2 interactors able to differentiate sPD and LRRK2-PD. λ value of 0.006, with log(λ) = −5.062 was selected according to 10-fold cross-validation. B LASSO coefficient profiles of 11 LRRK2 interactors are plotted. The optimal coefficient profile was produced against the selected λ (marked as the vertical red line). C The distribution curve shows different cut-off values and the model performance (as assessed by accuracy) on the train set. A cut-off of 0.54 was selected to reach the accuracy of 80.3%, with True Positive (TP) of 82.8% and True Negative (TN) of 57.9%. D The graph shows the ROC curve of the model validation on the test set = AUC value of 0.735.

Back to article page