Skip to main content
Fig. 6 | BMC Bioinformatics

Fig. 6

From: Interpretation of microbiota-based diagnostics by explaining individual classifier decisions

Fig. 6

Response to exclusive enteral nutrition (EEN) in Crohn’s disease patients: a clustered heat map displaying peaks’ relevance values across Crohn’s disease patients who received EEN treatment. Columns correspond to samples, rows to peaks (or bacterial species). For each sample, it is indicated whether the prediction was correct (i.e. the sample was classified as ‘IBD’, dark grey) or not (the sample was classified as ‘healthy, light grey), and whether remission was induced in response to EEN (green) or not (red). The cluster on the left is enriched with patients with induced remission compared to the other cluster on the right (p = 0.04, chi-square test). Only peaks with mean absolute relevance above 0.01 are shown. Color key from pink to blue indicates relevance from low to high, respectively. Clustering is based on cosine correlation matrix followed by unweighted pair group method with arithmetic mean (UPGMA). Relevance values were calculated based on linear SVM predictions

Back to article page