Additional file 3 of Location-specific signatures of Crohn’s disease at a multi-omics scale

Additional file 3: Supplementary Figure S3. Information supporting the distinction between ICD and CCD A) Comparison of mean abundance of Myosin-isoform proteins between ICD patients with either penetrating or structuring wounds (n = 10) to those with no wounds present (but still active disease, n = 9). * p < 0.05, Welch’s t-test. B) String-DB-generated network of combined inflammation-related cytokines and other proteins, filtered for high-confidence interactions (interaction score ≥ 0.7, see Supplementary Table 4) and proteins significantly increased in ICD compared to CCD. Grey dots represent imputed features, green is features increased in ICD. C). Expanded heatmap of features from the genus Blautia (x-axis) split by CD subtype (right) and further color coded by disease severity. See section 1 results for disease scoring split. Feature abundance scaled by column. D) Spearman correlation measurements comparing different bile acid families and SES-CD scores. Each dot represents mean correlation of a bile acid (or subtype) and metagenomic Blautia sp. features. E) Abundance of F. prausnitzii features from both metaproteome (left and middle) and ASV (right) feature sets in CD subtypes. * p < 0.05, *** p < 0.001, Welch’s t-test. F) ROC curve of individual features differentiating ICD and CCD. Features were chosen due to their statistical significance and representation of classes explored in the results. G) ROC curve generated by a model trained (ExtraTrees classifier) on all features in Supplementary Fig. 3F, and tested using a 70/30 split, using Leave-one-out cross validation.