Fig 1: Model performance in the validation cohort.a Comparison of AUROC for different proteomics-based models in predicting SI; b, c Comparison of balanced accuracy and F1-score across models for SI prediction; d Comparison of AUROC for different proteomics-based models in predicting 28-day mortality; e, f Comparison of balanced accuracy and F1-score across models for 28-day mortality prediction; g Confusion matrix for SI prediction; h Confusion matrix for 28-day mortality prediction; i Comparison of the expression levels of four selected proteins between patients with (n = 20) and without secondary infection (n = 40) in the validation cohort 1. For all statistical analyses, n denotes the number of biological replicates, defined as independent human subjects. Statistical analyses were performed by a two-tailed unpaired Student’s t test, and data were presented as mean ± SD. Statistical significance was observed for LYZ (P = 0.0005), SERPIND1 (P = 0.0030), CALM1 (P = 0.0295), and DPT (P = 0.0080). The center line denotes the median; box limits indicate the upper and lower quartiles; whiskers extend to 1.5× the interquartile range; points represent individual samples. Source data are provided as a Source Data file.