Leveraging sequence-based faecal microbial community survey data to identify a composite biomarker for colorectal cancer.

Abstract

Colorectal cancer (CRC) is the second leading cause of cancer-associated mortality in the USA. The faecal microbiome may provide non-invasive biomarkers of CRC and indicate transition in the adenoma-carcinoma sequence. Re-analysing raw sequence and metadata from several studies uniformly, we sought to identify a composite and generalisable microbial marker for CRC. Raw 16S rRNA gene sequence data sets from nine studies were processed with two pipelines, (1) QIIME closed reference (QIIME-CR) or (2) a strain-specific method herein termed SS-UP (Strain Select, UPARSE bioinformatics pipeline). A total of 509 samples (79 colorectal adenoma, 195 CRC and 235 controls) were analysed. Differential abundance, meta-analysis random effects regression and machine learning analyses were carried out to determine the consistency and diagnostic capabilities of potential microbial biomarkers. Definitive taxa, including <i>Parvimonas micra</i> ATCC 33270, <i>Streptococcus anginosus</i> and yet-to-be-cultured members of Proteobacteria, were frequently and significantly increased in stools from patients with CRC compared with controls across studies and had high discriminatory capacity in diagnostic classification. Microbiome-based CRC versus control classification produced an area under receiver operator characteristic (AUROC) curve of 76.6% in QIIME-CR and 80.3% in SS-UP. Combining clinical and microbiome markers gave a diagnostic AUROC of 83.3% for QIIME-CR and 91.3% for SS-UP. Despite technological differences across studies and methods, key microbial markers emerged as important in classifying CRC cases and such could be used in a universal diagnostic for the disease. The choice of bioinformatics pipeline influenced accuracy of classification. Strain-resolved microbial markers might prove crucial in providing a microbial diagnostic for CRC.