nPoRe: n-Polymer Realigner for improved pileup variant calling

Introduction

npore is a read realigner which recalculates each read's fine-grained alignment in order to more accurately align ''n-polymers'' such as homopolymers (n=1) and tandem repeats (2 ≤ n ≤ 6). In other words, given an input BAM, it adjusts each read's CIGAR string to more accurately model the most likely sequencing errors and actual variants. Traditional affine gap penalties are context-agnostic, and do not model the higher likelihood of INDELs in low-complexity regions (particularly n-polymers), leading to poor or inconsistent alignments. We find that npore improves pileup concordance across reads and results in slightly better variant calling performance.

Citation

Please cite the following pre-print if you use npore:

@article {dunn-npore,
    author = {Dunn, Tim and Blaauw, David and Das, Reetuparna and Narayanasamy, Satish},
    title = {nPoRe: n-Polymer Realigner for improved pileup variant calling},
    elocation-id = {2022.02.15.480561},
    year = {2022},
    doi = {10.1101/2022.02.15.480561},
    publisher = {Cold Spring Harbor Laboratory},
    URL = {https://www.biorxiv.org/content/early/2022/02/18/2022.02.15.480561},
    eprint = {https://www.biorxiv.org/content/early/2022/02/18/2022.02.15.480561.full.pdf},
    journal = {bioRxiv}
}

Contents

• Introduction
• Installation
• Usage
• Project Structure
• Data Sources
• Acknowledgements

Installation

Option 1: GitHub Source

First, clone the repository:

git clone https://github.com/timd1/npore && cd npore

Next, set up a virtual environment, activate it, and install the required packages.

python3 -m venv venv3 --prompt "npore"
source ./venv3/bin/activate
python3 -m pip install --upgrade pip
python3 -m pip install -r requirements.txt

Please ensure this environment is activated when building or running npore. Lastly, build npore and verify that it has succeeded.

make
python3 ./src/realign.py --help

Option 2: Docker Hub Image

A pre-built Docker image can be downloaded from here using:

sudo docker pull timd1/npore
sudo docker run -it timd1/npore:latest python3 realign.py --help

Option 3: Dockerfile

This may take some time to re-build the image; the previous option should be preferred in most cases.

git clone https://github.com/TimD1/npore && cd npore
sudo docker build -f ./Dockerfile -t timd1/npore:latest .
sudo docker run -it timd1/npore:latest python3 realign.py --help

Usage

Prerequisites

All input BAMs are required to:

1. Have MD tag annotations (for pysam read to ref mapping)
2. Be indexed (have an associated .bam.bai file)

You can prepare your input BAM using samtools:

samtools calmd -@ `nproc` -b -Q orig_reads.bam ref.fasta > reads.bam
samtools index reads.bam

Option 1: GitHub Source

Here's an example usage of the main realign.py program, which will store results in realigned.sam.

export NPORE="$HOME/npore"
export DATA="$NPORE/test/data"
. $NPORE/venv3/bin/activate
python3 $NPORE/src/realign.py \
    --bam $DATA/reads.bam \
    --ref $DATA/ref.fasta \
    --out_prefix $DATA/realigned \
    --stats_dir $NPORE/guppy5_stats

For additional options, run python3 realign.py --help.

Option 2: Docker

Here's how to call the Docker container with the same arguments as above:

export NPORE="$HOME/npore"
export DATA="$NPORE/test/data"
sudo docker run \
    -v $DATA:$DATA \
    -v $NPORE/guppy5_stats:$NPORE/guppy5_stats \
    timd1/npore:v0.1.0 \
        python3 realign.py \
        --bam $DATA/reads.bam \
        --ref $DATA/ref.fasta \
        --out_prefix $DATA/realigned \
        --stats_dir $NPORE/guppy5_stats

Project Structure

All other src/ files (aln.pyx, bam.pyx, cig.pyx, vcf.py, util.py) contain functions used in the above modules.

Data Sources

The Genome In A Bottle GRCh38 v4.1 ground truth VCF and benchmarking regions were downloaded from here. The GRCh38 human reference sequence and R9.4.1 reads basecalled with the Guppy 5.0.6 super-accuracy model were downloaded from ONT Open Datasets.

Acknowledgements

We would like to thank the developers of samtools, minimap2, pysam, clair3, pepper-deepvariant, pysankey, igv, and swalign. We would also like to thank GIAB and ONT for making their data available publicly.