Fig. 5

Constrained multivariate analysis using MixMC method (sparse partial least squares discriminant analysis) to differentiate patients based on smoking status (current or never) with respect to duodenal (2nd part) mucosa-associated microbiota profiles. Each point represents an individual patient, colour coded by smoking status. A “training set” which consisted of the FD and ID cohort was used to develop a discriminating model, which was then tested on the CD data set. The figure displays the resulting predicted classification of each sample, with closed circles representing the FD-ID cohort, and open triangles representing the CD cohort. Confidence ellipses (95%) are shown. Samples misclassified by the model are circled in red