Plasma meta-proteome and metabolome identify novel bacterial peptides and metabolites predictive of poor outcomes in Acute Liver Failure

Abstract Background and Aims: Acute liver failure (ALF) has high mortality, probably due to accumulation of microbial and metabolic factors in the circulation. Classification of such factors in plasma (non-invasive) could help in stratification of ALF patients predisposed to early mortality Method: Baseline plasma metabolomics and meta-proteomics were performed in 40 ALF patients and 5 healthy controls (training-cohort). The results were validated on test cohort; plasma and paired one drop blood; in 160 ALF patients using HRMS and machine learning (ML). Circulating indicator panel for non-survival prediction were identified and correlated with clinical parameters of ALF patients. Results: Metabolomic profile of ALF-Non-Survivors was distinct and showed significantly increase in bile-acid, tryptophan, tyrosine metabolism and others already known for inflammation, cell death and response to stress (p<0.01, FDR1.5). Plasma of ALF-Non-Survivors showed higher Alpha/Beta diversity (p<0.05) with increase in Proteobacteria, Firmicutes, Actinobacteria and others (p<0.05), functionally associated to energy, amino acids, xenobiotic metabolism when compared to ALF survivors (p<0.05). Interestingly, increase in bacterial-Taxa and functionality correlated with Chenodeoxycholic acid, 4-(2-Amino phenyl)-2,4-dioxobutanoate and L-Tyrosine and others metabolites in ALF-NS (R2>0.7, p<0.05). Specific increase of metabolite in ALF-NS such as L-Tyrosine, 4-(2-Amino phenyl)-2,4-dioxobutanoate, Chenodeoxycholic acid (linked to cell death and inflammation), Carnosine and alanyl-tyrosine (linked to negate oxidative stress) was directly correlated with clinical parameters (R2>0.85). POD of top 5 metabolites showed a diagnostic efficiency of 98% (AUC=0.98(0.92-1.0)) for mortality and validated in the validation cohort using five machine learning algorithms showed >98% accuracy/sensitivity/specificity for prediction of early mortality. Conclusion: Plasma microbiome and metabolome of ALF-Non-Survivors are distinct. Baseline increase in plasma metabolite signature panel could be used to segregate ALF patients predisposed to early mortality.