Fig. 4: Aggregated pathway contribution and feature importance differences across subpopulations. | Communications Medicine

Fig. 4: Aggregated pathway contribution and feature importance differences across subpopulations.

From: Interpretable machine learning leverages proteomics to improve cardiovascular disease risk prediction and biomarker identification

Fig. 4

a The aggregated contribution from top pathways predictive of CVD risk. b Models built based on a split by gender, age and BMI showed lower linear correlations of feature importance between the subgroup models compared with random splits shown as a dotted line (gender P = 2e-9, age P = 3e-4 and BMI P 0.10, two-sided Fisher’s Z-test). An asterisk denotes a significant difference, using a significance threshold of 0.05.

Back to article page