Data from: Distinguishing potential bacteria-tumor associations from contamination in a secondary data analysis of public cancer genome sequence data

Background: A variety of bacteria are known to influence carcinogenesis.
Therefore, we sought to investigate if publicly available whole genome and
whole transcriptome sequencing data generated by large public cancer
genome efforts, like The Cancer Genome Atlas (TCGA), could be used to
identify bacteria associated with cancer. The Burrows-Wheeler aligner
(BWA) was used to align a subset of Illumina paired-end sequencing data
from TCGA to the human reference genome and all complete bacterial genomes
in the RefSeq database in an effort to identify bacterial read pairs from
the microbiome. Results: Through careful consideration of all of the
bacterial taxa present in the cancer types investigated, their relative
abundance, and batch effects, we were able to identify some read pairs
from certain taxa as likely resulting from contamination. In particular,
the presence of Mycobacterium tuberculosis complex in the ovarian serous
cystadenocarcinoma (OV) and glioblastoma multiforme (GBM) samples was
correlated with the sequencing center of the samples. Additionally, there
was a correlation between the presence of Ralstonia spp. and two specific
plates of acute myeloid leukemia (AML) samples. At the end, associations
remained between Pseudomonas-like and Acinetobacter-like read pairs in
AML, and Pseudomonas-like read pairs in stomach adenocarcinoma (STAD) that
could not be explained through batch effects or systematic contamination
as seen in other samples. Conclusions: This approach suggests that it is
possible to identify bacteria that may be present in human tumor samples
from public genome sequencing data that can be examined further
experimentally. More weight should be given to this approach in the future
when bacterial associations with diseases are suspected.