Machine learning to identify potential biomarkers for sarcopenia in liver cirrhosis

Figure 1 Analysis of serum metabolic profiling between sarcopenic and non-sarcopenic cirrhotic patients. A: Orthogonal partial least squares discriminant analysis (OPLS-DA) of metabolites in positive mode; B: Volcano plot of metabolites by univariate analysis in positive mode; C: OPLS-DA of metabolites in negative mode; D: Volcano plot of metabolites by univariate analysis in negative mode; E: Metabolite classes with significantly different concentrations between sarcopenic and non-sarcopenic cirrhotic patients; F: Enriched metabolic pathways of differential metabolites. VIP: Variable importance in the projection.

Figure 2 Screening of candidate diagnostic biomarkers in sarcopenic cirrhotic patients. A: Coefficient profile plot of the Least Absolute Shrinkage and Selection Operator model for cirrhotic patients with sarcopenia showed the final parameter selection l (lambda); B: Top-10 biomarkers based on their discriminant ability in the Support Vector Machine-Recursive Feature Elimination algorithm; C: Top-10 biomarkers selected by using the random forest algorithm; D: The Venn diagram for four candidate metabolic biomarkers in sarcopenic cirrhotic patients by intersecting the results of three algorithms. LASSO: Least Absolute Shrinkage and Selection Operator; RF: Random Forest; SVM-RFE: Support Vector Machine-Recursive Feature Elimination.

Figure 3 Receiver operating characteristic analysis of potential biomarkers for differentiating cirrhotic patients with or without sarcopenia. A: N-Acetylcarnosine with an area under the receiver operating characteristic curve (AUC) of 0.8153 (95%CI: 0.7094–0.9213, P < 0.0001); B: 2-Stearylcitrate with an AUC of 0.7387 (95%CI: 0.6046–0.8728, P = 0.0022); C: CerP (d18:1/12:0) with an AUC of 0.7085 (95%CI: 0.57–0.8469, P = 0.0076); D: 3-Methyl-alpha-ionylacetate with an AUC value of 0.7468 (95%CI: 0.6099–0.8838, P = 0.0016); E: The combined model of four metabolites with an AUC of 0.9384.

Figure 4 Heatmap of the Spearman correlation coefficient matrix between selected metabolites and clinical variables. Red and blue indicate negative and positive correlations, respectively. Empty cells indicate non-significant correlations (P values > 0.05). ^aP < 0.05, ^bP < 0.01, ^cP < 0.001. BMI: Body mass index; Cr: Creatinine; SF: Serum ferritin.