Saboo K, Petrakov NV, Shamsaddini A, et al. - On machine learning (ML) analysis, stool microbiota composition relative to saliva are significantly more informative in distinguishing between controls and patients with cirrhosis and those with differing cirrhosis severity. Although logistic challenges exist, stool should be preferred over saliva for microbiome analysis.

• Alterations in salivary and stool microbiota are witnessed in cirrhosis.

• The capability of stool vs salivary microbiota in distinguishing between groups was assessed using ML.

• This study involved 269 participants; 87 controls and 182 cirrhosis cases, of which 57 had hepatic encephalopathy (HE), 78 were on PPI and 29 on rifaximin.

• Irrespective of the ML model, a significantly higher AUC in differentiating groups was offered by stool microbiota vs saliva.

• Individual microbiota driving the distinctions between controls vs cirrhosis were higher autochthonous taxa, oral-origin microbiota between PPI users and pathobionts and autochthonous taxa between rifaximin and HE cases.

• These were identified to be consistent with the core microbiome analysis findings.