Home

Awesome

Nanopolish

build and test

Software package for signal-level analysis of Oxford Nanopore sequencing data. Nanopolish can calculate an improved consensus sequence for a draft genome assembly, detect base modifications, call SNPs and indels with respect to a reference genome and more (see Nanopolish modules, below).

A note on R10 support

Presently nanopolish does not support R10.4 flowcells as variant and methylation calling is accurate enough to not require signal-level analysis. We intend to support signal exploration through eventalign but do not currently have a timeline for this as our development time is currently dedicated to other projects.

Release notes

Dependencies

A compiler that supports C++11 is needed to build nanopolish. Development of the code is performed using gcc-4.8.

By default, nanopolish will download and compile all of its required dependencies. Some users however may want to use system-wide versions of the libraries. To turn off the automatic installation of dependencies set HDF5=noinstall, EIGEN=noinstall, HTS=noinstall or MINIMAP2=noinstall parameters when running make as appropriate. The current versions and compile options for the dependencies are:

In order to use the additional python3 scripts within /scripts, install the dependencies via

pip install -r scripts/requirements.txt --user

Installation instructions

Installing the latest code from github (recommended)

You can download and compile the latest code from github as follows:

git clone --recursive https://github.com/jts/nanopolish.git
cd nanopolish
make

Installing a particular release

When major features have been added or bugs fixed, we will tag and release a new version of nanopolish. If you wish to use a particular version, you can checkout the tagged version before compiling:

git clone --recursive https://github.com/jts/nanopolish.git
cd nanopolish
git checkout v0.9.2
make

Nanopolish modules

The main subprograms of nanopolish are:

nanopolish call-methylation: predict genomic bases that may be methylated
nanopolish variants: detect SNPs and indels with respect to a reference genome
nanopolish variants --consensus: calculate an improved consensus sequence for a draft genome assembly
nanopolish eventalign: align signal-level events to k-mers of a reference genome

Analysis workflow examples

Data preprocessing

Nanopolish needs access to the signal-level data measured by the nanopore sequencer. The first step of any nanopolish workflow is to prepare the input data by telling nanopolish where to find the signal files. If you ran Albacore 2.0 on your data you should run nanopolish index on your input reads (-d can be specified more than once if using multiple runs):

# Index the output of the basecaller
nanopolish index -d /path/to/raw_fast5s/ -s sequencing_summary.txt basecalled_output.fastq # for FAST5 inout
nanopolish index basecalled_output.fastq --slow5 signals.blow5 # for SLOW5 input

The -s option tells nanopolish to read the sequencing_summary.txt file from Albacore to speed up indexing. Without this option nanopolish index is extremely slow as it needs to read every fast5 file individually. If you basecalled your run in parallel, so you have multiple sequencing_summary.txt files, you can use the -f option to pass in a file containing the paths to the sequencing summary files (one per line). When using SLOW5 files as the input (FAST5 can be converted to SLOW5 using slow5tools), -s option is not required and does not affect indexing performance.

Computing a new consensus sequence for a draft assembly

The original purpose of nanopolish was to compute an improved consensus sequence for a draft genome assembly produced by a long-read assembly like canu. This section describes how to do this, starting with your draft assembly which should have megabase-sized contigs. We've also posted a tutorial including example data here.

# Index the draft genome
bwa index draft.fa

# Align the basecalled reads to the draft sequence
bwa mem -x ont2d -t 8 draft.fa reads.fa | samtools sort -o reads.sorted.bam -T reads.tmp -
samtools index reads.sorted.bam

Now, we use nanopolish to compute the consensus sequence (the genome is polished in 50kb blocks and there will be one output file per block). We'll run this in parallel:

python3 nanopolish_makerange.py draft.fa | parallel --results nanopolish.results -P 8 \
    nanopolish variants --consensus -o polished.{1}.vcf -w {1} -r reads.fa -b reads.sorted.bam -g draft.fa -t 4 --min-candidate-frequency 0.1

This command will run the consensus algorithm on eight 50kbp segments of the genome at a time, using 4 threads each. Change the -P and --threads options as appropriate for the machines you have available.

After all polishing jobs are complete, you can merge the individual 50kb segments together back into the final assembly:

nanopolish vcf2fasta -g draft.fa polished.*.vcf > polished_genome.fa

Calling Methylation

nanopolish can use the signal-level information measured by the sequencer to detect 5-mC as described here. We've posted a tutorial on how to call methylation here.

To run using docker

First build the image from the dockerfile:

docker build .

Note the uuid given upon successful build. Then you can run nanopolish from the image:

docker run -v /path/to/local/data/data/:/data/ -it :image_id  ./nanopolish eventalign -r /data/reads.fa -b /data/alignments.sorted.bam -g /data/ref.fa

Credits and Thanks

The fast table-driven logsum implementation was provided by Sean Eddy as public domain code. This code was originally part of hmmer3. Nanopolish also includes code from Oxford Nanopore's scrappie basecaller. This code is licensed under the MPL.

The scripts/compare_methylation.py was originally provided in the example methylation data bundle which was obtained using:

curl -O warwick.s3.climb.ac.uk/nanopolish_tutorial/methylation_example.tar.gz
tar xvfz methylation_example.tar.gz
ls methylation_example/compare_methylation.py