The sensitivity of massively parallel sequencing for detecting candidate infectious agents associated with human tissue

Abstract

Massively parallel sequencing technology now provides the opportunity to sample the transcriptome of a given tissue comprehensively. Transcripts at only a few copies per cell are readily detectable, allowing the discovery of low abundance viral and bacterial transcripts in human tissue samples. Here we describe an approach for mining large sequence data sets for the presence of microbial sequences. Further, we demonstrate the sensitivity of this approach by sequencing human RNA-seq libraries spiked with decreasing amounts of an RNA-virus. At a modest depth of sequencing, viral transcripts can be detected at frequencies less than 1 in 1,000,000. With current sequencing platforms approaching outputs of one billion reads per run, this is a highly sensitive method for detecting putative infectious agents associated with human tissues.

Figure 1. Flow chart of subtraction methodology.

Paired end reads from a human sequence library are first filtered to remove low quality reads;  = 20 nt homopolymers), and reads comprised of artifactual adapter or primer sequences. Then, using BWA , read pairs are aligned to databases of human ribosomal sequences, transcript sequences, and genomic sequences. Remaining reads are then aligned sequentially to the genome, transcriptome and human rRNA using BWA . Reads that remain unaligned after comparison to the various human sequence databases are then aligned to a custom database (IAdb) of all known viral and bacterial complete genome sequences, using Novoalign (http://www.novocraft.com), with the requirement of correct pairing logic. Although not considered here, remaining reads can be characterized further by de novo assembly.

Figure 2. Circos plot detailing HaRNAV sequence recovery.

The red and blue lines represent reads aligning on the minus and plus strand, respectively. The Heterosigma akashiwo RNA virus has an 8,587 bp ss-RNA linear genome with a single CDS, shown in green on the circos plot. The read depth of coverage is shown in the centre of the plot. The genome is depicted by alternating black-white arcs of 500 bp in size.