NINJA-OPS: Fast Accurate Marker Gene Alignment Using Concatenated Ribosomes

Abstract

The explosion of bioinformatics technologies in the form of next generation sequencing (NGS) has facilitated a massive influx of genomics data in the form of short reads. Short read mapping is therefore a fundamental component of next generation sequencing pipelines which routinely match these short reads against reference genomes for contig assembly. However, such techniques have seldom been applied to microbial marker gene sequencing studies, which have mostly relied on novel heuristic approaches. We propose NINJA Is Not Just Another OTU-Picking Solution (NINJA-OPS, or NINJA for short), a fast and highly accurate novel method enabling reference-based marker gene matching (picking Operational Taxonomic Units, or OTUs). NINJA takes advantage of the Burrows-Wheeler (BW) alignment using an artificial reference chromosome composed of concatenated reference sequences, the “concatesome,” as the BW input. Other features include automatic support for paired-end reads with arbitrary insert sizes. NINJA is also free and open source and implements several pre-filtering methods that elicit substantial speedup when coupled with existing tools. We applied NINJA to several published microbiome studies, obtaining accuracy similar to or better than previous reference-based OTU-picking methods while achieving an order of magnitude or more speedup and using a fraction of the memory footprint. NINJA is a complete pipeline that takes a FASTA-formatted input file and outputs a QIIME-formatted taxonomy-annotated BIOM file for an entire MiSeq run of human gut microbiome 16S genes in under 10 minutes on a dual-core laptop.

Patterns of seasonality and group membership characterize the gut microbiota in a longitudinal study of wild Verreaux's sifakas (Propithecus verreauxi)

“…We performed closed‐reference OTU‐picking and assigned taxonomy using the NINJA‐OPS classifier (Al‐Ghalith, Montassier, Ward, & Knights, 2016) with an 87% similarity threshold against the Greengenes 94% representative sequence database. We used the 87% threshold, which is the minimum generalized threshold for family‐level taxonomy discrimination (Yarza et al., 2014), because despite the high quality of the sequences as validated by quality scores, read lengths, and BLAST tests on a randomized subset of 1,000 reads, only approx.…”

Urinary microbiome associated with chronic allograft dysfunction in kidney transplant recipients

“…Data were quality controlled and stitched using the NINJA‐SHI7 fastq‐to‐fasta pipeline . Operational taxonomic units (OTUs) were selected using the NINJA‐OPS pipeline and the Greengenes for bacterial reference database, and RTL and UNITE for fungal reference databases (using a threshold of 97% similarity). Further taxonomic compositional and diversity analyses were performed using QIIME 1.9.1 …”

BugBase predicts organism-level microbiome phenotypes

“…For the Yellowstone data, the resulting OTU tables and mapping files were parsed to include only water samples (n=412). Sequences from the vaginal microbiome study [25] were downloaded from the National Center for Biotechnology Information Short Read Archive (SRA022855), and OTUs were picked using NINJA-OPS [41] and a closed reference approach at 97% identity against the 97% OTU Greengenes representative set (v13.5.8) [37]. Samples with at least 500 sequence counts were retained for the final anlysis (low Nugent score, n =248; intermediate Nugent score, n=49; high Nugent score, n= 97).…”