Fig. 1: Study design and alterations in the serum proteome between children with cerebral palsy and healthy controls.

A The proteomics workflow outlining three modules: sample collection, LC/MS analysis, and data analysis. B Pie chart showing clinical information of CP samples, including clinical subtypes, MRI, gender, GMFCS, and birth weight. C Heatmap and Bubble plot illustrating distinct proteins and pathways differentiating CP (n = 346) from HC (n = 190). The bubble size corresponds to the significance of the pathways enriched in each group. The red bubbles indicate CP, and the blue bubbles indicate HC. The two-sided Wilcoxon rank-sum test was used for differential analysis, with p-values adjusted for FDR. D The schematic showing the machine learning framework used to develop classifiers for identifying CP and healthy control samples. E The feature importance in distinguishing CP from healthy controls using the XGBoost-based model. F The Receiver operating characteristic (ROC) curve showing the performance of the multi-biomarker model in the testing set based on proteomic results, and the confusion matrix showing the performance of the ML classifier. G Boxplots showing 10 diagnostic protein differences between CP (n = 38) and healthy control (n = 32) in the validation cohort using ELISA (two-sided Wilcoxon rank-sum test). Boxplots show median (central line), upper and lower quartiles (box limits), 1.5× interquartile range (whiskers). H The ROC curve showing the performance of the multi-biomarker model in the testing set based on ELISA results, and the confusion matrix showing the performance of the ML classifier. Source data are provided as a Source Data file.