Bayesian mixture analysis for metagenomic community profiling.

Sofia Morfopoulou; Vincent Plagnol

We followed up on the sequences assigned to the "unknown category", that is approximately 170K reads, looking for nucleotide similarity with NR-NT using BLASTn. Approximately half of the reads originated from an untranslated region of the Coronavirus genome, which is not captured by the protein reference database. The remaining reads matched confidently to either Danio rerio (zebrafish) sequences or Gallus gallus (chicken), two organisms whose proteins are not in the human microbiome reference we are using. The zebrafish and chicken matches were explained as barcode leakage resulting from multiplexing on the same flowcell zebrafish and chicken RNA-Seq libraries. metaMix appropriately assigned these reads to the "unknown" category, producing a clean probabilistic summary (Supplementary Table S2 ). The method ran in 4.7 hours.