P855 Variable importance analysis based on gut microbiota and dietary factors between IBD patients and healthy controls in China

Abstract

Background: Gut microbiota and diet are believed to be associated with the pathogenesis and development of inflammatory bowel disease (IBD). Our study investigated the differences in gut microbiota and dietary factors between Chinses IBD patients and their cohabitating family member controls Methods: We recruited Crohn's disease (CD) and ulcerative colitis (UC) patients with endoscopically confirmed disease from the IBD clinic in 6th Affiliated Hospital of Sun Yat-sen University in Guangzhou, China between March 2014 and September 2016. Each case was asked to provide a family member (primarily sibling) control. Individuals who had not taken antibiotics in the prior 2 weeks provided stool samples with 24 h dietary recalls. Physicians completed information to calculate the Mayo and CDAI scores using the most recent laboratory and endoscopic information. Faecal bacterial differential diversity were analysed with Miseq sequencing results of the V5-V6 region of the 16S rDNA. The Wilcoxon signed-rank test was used to make taxonomy-based comparisons of gut microbiota for 200 differential operational taxonomic units (OTUs) selection including levels of family, genus and species. Dietary records were entered and computed with the NCI Automated Self-Administered 24-h Dietary Assessment Tool and total energy (Kcal) adjusted protein, sugar, fibre, total monounsaturated fatty acids, and total saturated fatty acids were calculated. Statistically significant OTU and dietary factors were selected with Random Forests classifier using R selecting the top 200 factors using univariable inputs. Results: Stool-dietary recall case-control paired results were available for 37 CD and 14 UC patients and their matched controls (n = 51). The median age at the time of stool collection was 29 in cases and 30 in controls. According to Mayo and CDAI scoring systems, 65% of UC and 70.7% of CD cases were in remission. No case had severe disease at the time of stool collection. The differences in out profiles were statistically significant for CD cases compared with controls (p < 0.04), but not UC (p < 0.17). Comparison between UC and CD pairs showed that only 2 OTUs had a similar distribution. When we examined univariable factors, contained both 16S rDNA and nutrition data, that differentiated cases and controls, Lachnospiraceae and Rumino-coccaceae were the most important families. Dietary factors ranked below the top 50 with monosaturated fat (rank 59) and protein (rank 67) the highest rated. Fibre ranked 158 out of 200. Conclusions: Microbiota profiles are more important than dietary factors to differentiate IBD from controls, especially for CD.