Fig. 3: Differential analysis and classification with MRM-identified DEPs and clinical indexes.

Differential analysis and machine learning were performed with the identified proteins and clinical indexes in four comparison groups (pathogenesis: AMS4k (n = 30) and AMS1k (n = 30); protection: nAMS4k (n = 23) and nAMS1k (n = 23); prediction: AMS1k (n = 30) and nAMS1k (n = 23); and diagnosis: AMS4k (n = 30) and nAMS4k (n = 23). a Heatmap of the log₂FC of 75 PEA- or MRM-identified DEPs in the 5 comparison groups based on PEA or MRM. b Heatmap of the log₂FC of 37 differential clinical indexes involved in the 4 comparison groups. The heatmap depicts the regulation trends of each protein or clinical index with the legend at the left of the figure. The statistical methods were used paired/unpaired t test or Welch’s t test. The differentially expressed proteins and clinical indexes are labeled with * (q value < 0.05) and +(p value < 0.05). c ROCs of the classification models with proteins (orange curve), clinical indexes (blue curve), or combined data (yellow curve) and bar plot of important proteins (orange bar) between AMS4k and AMS1k (pathogenesis, pink box). All the pathogenesis models with three types of features exhibited high accuracy with statistical significance, as demonstrated by the AUCs and p values. d ROCs of the classification models with proteins (orange curve), clinical indexes (blue curve), or combined data (yellow curve) and bar plot of important proteins (orange bar) between the nAMS4k and nAMS1k groups (protection, dark blue box). All protection models with three types of features show high accuracy with statistical significance. e ROCs of the classification models with proteins (orange curve), clinical indexes (blue curve), or combined data (yellow curve) and bar plot of important proteins (orange bar) between the AMS1k and nAMS1k groups (prediction, dark green box). The prediction models with proteins and combined data show high accuracy. f ROCs of the classification models with proteins (orange curve), clinical indexes (blue curve), or combined data (yellow curve) and bar plot of important proteins (orange bar) between the AMS4k and nAMS4k groups (diagnosis, yellow box). The diagnosis models with proteins and combined data exhibit high accuracy with statistical significance. The p value (in parentheses) of each AUC is shown after the AUC value. ROC receiver operating characteristic curve, AUC area under the curve.