IDDF2024-ABS-0442 A predictive machine learning model based on gut microbial signatures for clinical remission in inflammatory bowel disease treated with biologics: a comprehensive multi-cohort analysis

<h3>Background</h3> The onset and progression of inflammatory bowel disease (IBD) are closely associated with gut microbiota. Emerging evidence suggests a potential link between gut microbiota and the effectiveness of biologics in IBD. However, the specific gut microbial features that correlate with the effectiveness of biologics remain poorly defined. This study aims to identify specific gut microbial signatures for clinical remission (CR) in IBD patients treated with biologics and to develop a predictive machine-learning model based on these signatures. <h3>Methods</h3> This meta-analysis included 2331 stool samples of 16S rRNA gene sequencing data from two independent studies (231 patients). All the participants were treated with biologics: 23 with infliximab, 22 adalimumab and 186 ustekinumab. Alterations in diversity and taxonomic compositions of gut microbiota were compared between patients with CR and those without (non-CR). A machine learning model was developed to predict CR based on gut microbial signatures. <h3>Results</h3> Principal coordinate analysis showed significant differences in beta diversity between CR and non-CR patients across cohorts (<i>P</i> &lt; 0.05), while no significant differences in alpha diversity were found (<i>P</i> &gt; 0.05). Forty-one differential genera were identified between CR and non-CR patients. Notably, genera <i>Parabacteroides B 862066, Lachnoclostridium B, Faecalibacterium, Faecousia, Onthenecus, SFMI01, Butyribacter</i>, and <i>Merdibacter</i> were significantly decreased (P &lt; 0.05) in the non-CR group, while <i>UMGS1375</i> and <i>Anaeroglobus</i> were significantly increased. The prediction model, trained by the top 10 differential genera, effectively discriminates between patients achieving CR and non-CR with an area under the receiver operating characteristic curve of 0.80. <h3>Conclusions</h3> This study is the first to demonstrate that IBD treatment outcomes are associated with distinct gut microbiota profiles across multiple cohorts. Significant changes in gut microbiota are associated with CR in patients treated with biologics. A machine learning model based on gut microbial signatures effectively predicted CR. Further studies are needed to explore the potential benefits of modulating the gut microbiome to improve the efficacy of biological treatments for IBD.