Fig. 5

Random forest classifications performed on 16 artificial bacterial assemblages generated in silico. Red dots show the proportion of sequences within the total assemblage belonging to a source (1 dot = 1%). The percentage listed in the freshwater column corresponds to the proportion of sequences from the non-contaminated freshwater sample within the total assemblages (per sample, red dots + freshwater percentages = 100%). Bold values associated with red dots indicate the proportion of contamination expected for the different sources. Predictions of the Clostridiales and Bacteroidales classifications are indicated in the white rows. Green circles indicate the classifier detected the source signature in the sample. Orange circles indicate the classifier did not detect the source signature when it was expected. Blue circles indicate the classifier did not detect a signature when a source not included in the classifiers was included in the assemblage. The proportion of sequences matching the source classifier is associated with the green circles. See Additional file 2 for more details