Fig. 3
From: CAPN2 correlates with insulin resistance states in PCOS as evidenced by multi-dataset analysis
Machine Learning for Pivotal Gene Selection. (A) LASSO regression is shown via coefficient trajectories against log(lambda) on the left, identifying optimal gene selection parameters: ‘Lambda.min’ for minimal error and ‘Lambda.1se’ for a simpler model. On the right, a misclassification error plot indicates the best-performing model against lambda values for robust PCOS gene selection. (B) The RFE analysis correlates the number of features with model accuracy (left) and cross-validation error (right). The plots highlight the optimal number of predictive features that correspond to the highest accuracy and lowest error, optimizing the gene selection for PCOS. (C) Boruta’s boxplot assesses gene importance against shadow features to confirm the significance of genes in PCOS, distinguishing those with substantial contributions to the pathogenesis of IR.