Fig. 2

Influences of THBS1 on HBV-ACLF prognosis and short-term mortality. A Human PBMC transcriptome expression profile of THBS1 in HBV-ACLF deceased group (n = 8/9) was significantly higher than HBV-ACLF survival group (n = 12/11) at 28/90 days. B PCR validation in human PBMCs confirmed the significant difference between HBV-ACLF deceased group (n = 30/38) and HBV-ACLF survival group (n = 60/44) at 28/90 days, * P < 0.05, ** P < 0.01. Twelve patients with ACLF underwent liver transplantation and were considered lost to follow-up in the mortality rate calculation. Six patients with ACLF were lost to the 28-day follow-up analysis, and 16 patients were lost to the 90-day follow-up analysis. C Plasma THBS1, measured by ELISA, verified the significant difference between HBV-ACLF deceased group (n = 66/91) and HBV-ACLF survival group (n = 132/107) at 28/90 days, *** P < 0.001. D, E Accuracy of THBS1 in predicting ACLF short-term mortality was evaluated through the area under the receiver operating characteristic curve (AUROC) in the sequencing group (D), and plasma THBS1 in the validation group (E) for 28 and 90 days. F Kaplan‒Meier curves of survival probability stratified by plasma THBS1 level at 28 days (left) and 90 days (right). Level of significance: P < 0.0001 (log-rank test). The number of patients at risk is shown for each observation period. G The correlation between serum THBS1 and the clinical parameters ALT and γ-GT (n = 198). H Correlations between THBS1 expression in PBMCs and the different scoring systems (n = 98); THBS1 was significantly (P < 0.05) positively correlated with COSSH-ACLFs, MELD-Nas, MELDs and CTP, whereas no significant (P > 0.05) correlation existed with either COSSH-ACLF IIs or CLIF-C ACLFs. I The correlation between THBS1 levels and selected inflammatory mediators (grey) and apoptotic targets (pink) in the human PBMC transcriptome in the four groups comparison (ACLF vs. LC vs. CHB vs. NC). Pearson’s correlation coefficient (r) was used for the assessment of the correlations, and a correlation plot was generated using the R corrplot package