DecontaMiner: A Pipeline for the Detection and Analysis of Contaminating Sequences in Human NGS Sequencing Data

Abstract

Reads alignment is an essential step of next generation sequencing) data analyses. One challenging issue is represented by unmapped reads that are usually discarded and considered as not informative. Instead, it is important to fully understand the source of those reads, to assess the quality of the whole experiment. Moreover, it is of interest to get some insights on possible “contamination” from non-human sequences (e.g., viruses, bacteria, and fungi). Contamination may take place during the experimental procedures leading to sequencing, or be due to the presence of microorganisms infecting the sampled tissues. Here we propose a pipeline for the detection of viral, bacterial, and fungi contamination in human sequenced data. Similarities between input reads (query) and putative contaminating organism sequences (subject) are detected using a local alignment strategy (MegaBLAST). For each organism database DecontaMiner provides two main output files: one containing all the reads matching only a single organism; the second one containing the “ambiguous” matching reads. In both files, data is sorted by organism and classified by taxonomic group. Low quality, unaligned sequences, and those discarded by user criteria are also provided as output. Other information and summary statistics on the number of matched/filtered/discarded reads and organisms are generated. This pipeline has successfully detected foreign sequences in human Cancer RNA-seq data.