nhoffman / 16s-capture

Pipleine code for "Sensitive identification of bacterial DNA in clinical specimens by broad range 16S rRNA enrichment"

16S rRNA Capture Pipeline

Pipleine code for "Sensitive identification of bacterial DNA in clinical specimens by broad range 16S rRNA enrichment"

Sara Rassoulian Barrett [1], Noah G. Hoffman [1], Christopher Rosenthal [1], Andrew Bryan [1], Desiree A. Marshall [2], Joshua Lieberman [1], Brad T. Cookson [1], Stephen J. Salipante [1] (corresponding author)

• [1] Department of Laboratory Medicine, University of Washington, Seattle, WA, USA
• [2] Department of Pathology, University of Washington, Seattle, WA, USA

setup

create a virtualenv:

python3 -m venv py3-env
source py3-env/bin/activate
pip install -U pip wheel
pip install scons==3.0.5
pip install -r requirements.txt

Extract PEAR binary: PEAR 0.9.11 binaries and sources were downloaded from http://www.exelixis-lab.org/web/software/pear - this required filling out a registration form that worked only on Chrome. These are stored in /src. To copy the binary to the virtualenv:

tar -xf src/pear-0.9.11-linux-x86_64.tgz --strip-components 2 -C py3-env/bin pear-0.9.11-linux-x86_64/bin/pear

Build the Singularity images:

mkdir -p singularity_images
for fn in singularity/*; do sudo singularity build singularity_images/${fn#singularity/Singularity_}.simg $fn; done

methods

Reference set creation

A bacteria mock community of 16S rRNA gene reference sequences was acquired and assembed from BEI Resources and aligned and used to create phylogenetic trees [1]. Two additional reference packages were assembled by recruiting 16S rRNA reference sequences from a ya16sdb 0.4 curated set of NCBI 16s sequences [2] and selecting based on similarity to clinical specimens using DeeNuRP 0.2.4 search-sequences and select-references [3][4].

16s Classification

Illumina MiSeq reads were filtered, trimmed, deduplicated and assembled using barcodecop 0.5 [5], ea-utils 1.04.807 fastqc-mcf [6], HTStream 0.3.0 SuperDeduper [7] and PEAR 0.9.11 [8] respectively. 16s reads were selected using Infernal 1.1.2 cmsearch and aligned using cmalign [9]. The resulting alignments were merged with reference alignments (using 'esl-alimerge') to place all sequences in the same alignment register. Query sequences were then placed on a phylogenetic tree of reference sequences using epa-ng 0.3.5 [10] and classified using gappa 0.2.4 [11]. The full Python Scons pipeline is available for evalutation at https://github.com/salipante/16s-capture.git.

1. bei resources
2. https://github.com/nhoffman/ya16sdb
3. https://github.com/fhcrc/deenurp/tree/master/deenurp
4. https://doi.org/10.1371/journal.pone.0065226
5. https://github.com/nhoffman/barcodecop
6. https://github.com/ExpressionAnalysis/ea-utils
7. https://github.com/ibest/HTStream
8. https://doi.org/10.1093/bioinformatics/btt593
9. https://doi.org/10.1093/bioinformatics/btt509
10. https://doi.org/10.1093/sysbio/syy054
11. https://doi.org/10.1093/bioinformatics/btaa070