Establishing a novel colorectal cancer predictive model based on unique gut microbial single nucleotide variant markers
Chenchen Ma, Kaining Chen, Yuanyuan Wang, Chaoping Cen, Qixiao Zhai, Jiachao Zhang
Abstract
Current metagenomic species-based colorectal cancer (CRC) microbial biomarkers may confuse diagnosis because the genetic content of different microbial strains, even those belonging to the same species, may differ from 5% to 30%. Here, a total of 7549 non-redundant single nucleotide variants (SNVs) were annotated in 25 species from 3 CRC cohorts (n = 249). Then, 22 microbial SNV markers that contributed to distinguishing subjects with CRC from healthy subjects were identified by the random forest algorithm to construct a novel CRC predictive model. Excitingly, the predictive model showed high accuracy both in the training (AUC = 75.35%) and validation cohorts (AUC = 73.08%-88.02%). We further explored the specificity of these SNV markers in a broader background by performing a meta-analysis across 4 metabolic disease cohorts. Among these SNV markers, 3 SNVs that were enriched in CRC patients and located in the genomes of Eubacterium rectale and Faecalibacterium prausnitzii were CRC specific (AUC = 72.51%-94.07%).