Fig. 3
The log10-transformed ratios of bacterial read pairs are shown for all nine cancer types comparing counts of Mycobacterium (M), Acinetobacter (A), Pseudomonas (Ps), Staphylococcus epidermidis (S), Ralstonia (R), and Propionibacterium (P) read pairs. Staphylococcus and Propionibacterium were grouped together after an initial comparison found them to frequently have the same relative proportions. In addition, they are both commonly found on the human skin and may arrive in the samples through the same mechanism. The black horizontal lines represent the average of the log10-transformed ratios across all the cancer types. When points are clustered near the mean from all cancer types, it suggests a common source of the bacterial reads. Given the diverse cancer types and the numerous collection and sequencing centers, we interpret those contributions to be from a general source of contamination, whereas when a set of samples does not cluster with the others (e.g., OV and GBM in M/R), it suggests a more specific source of the bacterial sequences. The latter can be due either to contamination or a biological reason, which cannot be distinguished here. This was found for comparisons containing (a) Mycobacterium in OV and GBM, (b) Pseudomonas in STAD, and (c) Acinetobacter, Ralstonia, and Pseudomonas in AML. The standard deviations with all data, depicted by the vertical black line, of each of the bacterial comparisons are shown across all nine datasets. AML, OV, and STAD, which are cancer types that have at least one predominant bacterial species, not including Enterobacteriaceae, were excluded from a subsequent mean and standard deviation calculation, depicted by the gray horizontal and vertical lines, respectively. Excluding AML, OV, and STAD decreases the standard deviation for the comparisons involving Acinetobacter spp. and Pseudomonas spp. This suggests that the presence of Acinetobacter spp. and Pseudomonas spp. may be attributed to similar levels of general contamination of all samples by a similar mechanism, except in AML, OV, and STAD