Machine learning identifies SPP1-related modules as top predictive features (IMAGE)
Caption
(A) Bar plot showing the top 15 feature importance scores from Random Forest classification, with SPP1-related genes among the top predictors. (B) Confusion matrix demonstrating classification performance between predicted and true labels (0 = Sham, 1 = Operation). (C) Receiver operating characteristic (ROC) curve showing excellent model performance (AUC = 0.99). (D) Learning curves showing training and validation loss across epochs, demonstrating stable convergence without overfitting. (E) SHAP summary plot showing feature contributions to model predictions, with each point representing a sample and color indicating the feature value (red = high, blue = low). (F) SHAP feature importance bar plot showing mean absolute SHAP values for the top features, including Fcer1g, Tyrobp, Lyz2, Chondrocyte, Serpinb1a, Gpr183, and RT1-Da. (G) PCA of 1423-dimensional data showing separation between groups. (H) Score plot with gradient colors demonstrating principal component distribution. (I, J) K-Means clustering results (Optimal K = 5) in 2D and 3D representations with centroids marked.
Credit
Qiuwei Li, Liang Kang, Kaida Bo, Peilin Jin, Chenhao Zhao, Renjie Zhang, Cailiang Shen
Usage Restrictions
Credit must be given to the creator. Only noncommercial uses of the work are permitted. No derivatives or adaptations of the work are permitted.
License
CC BY-NC-ND