A Snakemake wrapper for easily creating de novo bacterial isolate genome assemblies from Oxford Nanopore (ONT) sequencing data, and optionally Illumina data, using any combination of read filtering, assembly, long and short read polishing, and reference-based polishing.
The workflow is published in Snakemake workflows for long-read bacterial genome assembly and evaluation in GigaByte.
read filtering | assembly | long read polishing | short read polishing | reference-based polishing |
---|---|---|---|---|
Filtlong Rasusa |
Flye raven miniasm Unicycler Canu |
racon medaka |
pilon Polypolish POLCA |
Homopolish proovframe |
# Install
git clone https://github.com/pmenzel/ont-assembly-snake.git
conda config --add channels bioconda
conda env create -n ont-assembly-snake --file ont-assembly-snake/env/conda-main.yaml
conda activate ont-assembly-snake
# Prepare ONT reads, one file per sample
mkdir fastq-ont
cp /path/to/my/data/my_sample/ont_reads.fastq.gz fastq-ont/mysample.fastq.gz
# optionally: add Illumina paired-end reads
mkdir fastq-illumina
cp /path/to/my/data/my_sample/illumina_reads_R1.fastq.gz fastq-illumina/mysample_R1.fastq.gz
cp /path/to/my/data/my_sample/illumina_reads_R2.fastq.gz fastq-illumina/mysample_R2.fastq.gz
# Declare desired combination of read filtering, assembly and polishing
mkdir assemblies
mkdir assemblies/mysample_flye+medaka
mkdir assemblies/mysample+filtlongMB500_flye+racon2+medaka
mkdir assemblies/mysample_raven2+medaka+pilon
[...]
# Run workflow
snakemake -s ont-assembly-snake/Snakefile --use-conda --cores 20
Clone repository, for example into the existing folder /opt/software/
:
git clone https://github.com/pmenzel/ont-assembly-snake.git /opt/software/ont-assembly-snake
Install conda and create a new environment called ont-assembly-snake
:
conda config --add channels bioconda
conda env create -n ont-assembly-snake --file /opt/software/ont-assembly-snake/env/conda-main.yaml
Activate the environment:
conda activate ont-assembly-snake
First, prepare a folder called fastq-ont/
containing the ONT sequencing reads as
one .fastq
or .fastq.gz
file per sample, e.g. fastq-ont/sample1.fastq.gz
.
Unicycler and some polishing tools additionally use paired-end Illumina reads, which need to be placed in the folder fastq-illumina/
using _R[12].fastq
or _R[12].fastq.gz
suffixes, e.g. fastq-illumina/sample1_R1.fastq.gz
and fastq-illumina/sample1_R2.fastq.gz
.
Next, create a folder assemblies
and in there, create empty folders specifying
the desired combinations of read filtering, assembly, and polishing steps by using specific keywords for each program, see below.
The first part of a folder name is a sample name, which must match the filenames in fastq-ont/
and, optionally, fastq-illumina/
.
The next part can be a keyword for read filtering with Filtlong, see below, which is separated from the sample name by +
.
Then follows, separated by an underscore, a keyword for the assembler.
NB: This also means that sample names must not contain underscores.
After the keyword for the assembler follow the keywords for one ore more polishing steps, all separated by +
.
As an alternative to create subfolders within assemblies/
, it is also possible to use a YAML file to list the desired assemblies, e.g. in a file called samples.yaml
:
assemblies:
- mysample_flye+medaka
- mysample+filtlongMB500_flye+racon2+medaka
- mysample_raven2+medaka+pilon
and add the argument --configfile samples.yaml
to the Snakemake command line.
Finally, after making the desired subfolders in assemblies/
or making the config file, run the workflow, e.g. with 20 threads:
snakemake -s /opt/software/ont-assembly-snake/Snakefile --use-conda --cores 20
The assemblies created in each step are contained in the files output.fa
(FASTA format) in each of the previously created folders within assemblies/
.
Additionally, a symlink named <FOLDER_NAME>.fa
links to the final output file of each step, i.e. assemblies/<FOLDER_NAME>/output.fa
, also see the example below.
The log files of running the individual Snakemake rules are located in assemblies/<FOLDER_NAME>/*log.txt
.
A test dataset containing a pair of ONT and Illumina sequencing data from the same bacterial isolate is available in the repository ont-assembly-snake-testdata. See the instructions therein on how to download the dataset and run the ont-assembly-snake and score-assemblies workflows.
The ONT reads can be filtered by length and quality using Filtlong prior to the assembly.
The available keywords are:
filtlong
:
This will filter the ONT reads in fastq-ont/mysample.fastq
and keep only
reads longer than 1000 bases; using the Filtlong option --min_length
. The filtered read set is written to
fastq-ont/mysample+filtlong.fastq
. The length can be changed using the
Snakemake configuration option filtlong_min_read_length
.
filtlongPC<p>
This will filter the reads to only include the top p
percent of megabases from reads with highest average quality using
the Filtlong option --keep_percent
.
Further, reads are filtered by their length as above. The output is written to fastq-ont/mysample+filtlongPC<p>.fastq
.
filtlongMB<m>
This will filter the reads to only include reads with highest average quality up to a total length of m
megabases.
Further, reads are filtered by their length. The output is written to fastq-ont/mysample+filtlongMB<m>.fastq
.
filtlongMB<m>,<q>,<l>
This will filter the reads to only include reads up to a total length of m
megabases, which are filtered by length
and quality, where q
and l
set the priority for each using the Filtlong options --mean_q_weight
and --length_weight
, respectively.
See also the section in the Filtlong docs.
Further, reads are filtered by their length as above. The output is written to fastq-ont/mysample+filtlongMB<m>,<q>,<l>.fastq
.
filtlongMB<m>,<q>,<l>,<n>
As above, but the the minimum read length is explicitly specified by n
and not by the global option filtlong_min_read_length
:
The output is written to fastq-ont/mysample+filtlongMB<m>,<q>,<l>,<n>.fastq
.
When using any of the Filtlong keywords in a folder name, they must be followed by an underscore, followed by the keyword for the assembler.
The ONT reads can be randomly subsampled prior to the assembly.
The available keywords are:
rasusaMB<m>
This will subsample the ONT reads to a total of m
megabases.
The output is written to fastq-ont/mysample+rasusaMB<m>.fastq
.
When using any the Rasusa keyword in a folder name, it must be followed by an underscore, followed by the keyword for the assembler.
Following keywords can be used to run the assembly with Flye:
flye
Default assembly, which includes one round of internal polishing the assembly with the ONT reads.
flyeX
Assembly with X
rounds of internal polishing. Setting X
to 0 disables polishing altogether.
flyehq
Assembly for high-quality ONT reads using Flye option --nano-hq
for ONT Guppy5+ in SUP mode, with one round of internal polishing.
flyehqX
High-quality assembly, with X
rounds of internal polishing. Setting X
to 0 disables polishing altogether.
Following keywords can be used to run the assembly with raven:
raven
Default assembly, which includes two rounds of internal polishing with racon using the ONT reads.
ravenX
Assembly with X
rounds of internal polishing with racon. Setting X
to 0 disables polishing altogether.
Following keywords can be used to run the assembly with miniasm:
miniasm
Default assembly. Miniasm does not do any polishing by itself.
unicycler
Unicycler does a hybrid assembly, i.e., both ONT and Illumina reads must be present in fastq-ont
and fastq-illumina
, respectively.
canu
The Canu assembler requires to know the genome size (in Megabases) beforehand, use Snakemake option: --config genome_size=5.2
(e.g. for 5.2 Mb)
NB: Canu can take a long time to complete the assembly, up to several hours!
Following keywords can be used to polish an assembly using ONT reads:
racon
Polishing the assembly once.
raconX
Run racon polishing iteratively X
times.
medaka
Medaka polishes the assembly using the ONT reads, but also requires the name of
the Medaka model to be used, which depends on the flow cell and basecalling that were used for creating the reads.
The model name can either be set globally for all samples using the Snakemake configuration option medaka_model
,
or by supplying a tab-separated file with two columns that maps sample names to medaka models using the Snakemake configuration option map_medaka_model
.
Options are specified using snake make's --config
parameter, for example:
snakemake /opt/software/ont-assembly-snake/Snakefile --cores 20 --config map_medaka_model=map_medaka.tsv
where map_medaka.tsv
contains, for example, the two columns:
sample1 r941_min_high_g330
sample2 r941_min_high_g351
If no model is specified, Medaka will use the r941_min_hac_g507
model by default.
pilon
Pilon polishes an assembly using Illumina reads, which must be located in the fastq-illumina
folder.
polypolish
Polypolish polishes an assembly using Illumina reads, which must be located in the fastq-illumina
folder.
polca
POLCA polishes an assembly using Illumina reads, which must be located in the fastq-illumina
folder.
homopolish
Homopolish does reference-based polishing based on one ore more provided reference genomes in FASTA format located in
references/NAME1.fa
, references/NAME2.fa
, etc., where NAME1
and NAME2
can be any string.
Snakemake will create output files ...+homopolish/output_NAME1.fa
, ...+homopolish/output_NAME1.fa
, etc., containing the polished assemblies.
When using Homopolish, it must be the last keyword in the folder name.
proovframe
Proovframe does reference-based polishing based on one ore more provided reference proteomes in FASTA format containing the amino acid sequences located in
references-protein/NAME1.faa
, references-protein/NAME2.faa
, etc., where NAME1
and NAME2
can be any string.
Snakemake will create output files ...+proovframe/output_NAME1.fa
, ...+proovframe/output_NAME1.fa
, etc., containing the polished assemblies.
When using proovframe, it must be the last keyword in the folder name.
This example contains two samples with ONT sequencing reads and Illumina reads for sample 2 only.
For sample 1, the assembly should be done with Flye (including the default single round of
polishing), followed by polishing the assembly with racon (twice), medaka, and eventually Homopolish, which will use the E. coli genome in the file references/Ecoli.faa
.
In another assembly, we also want to filter the ONT reads of sample 1 to only include the highest quality reads up to a total of 500Mb
using Filtlong and apply the same assembly and polishing protocol.
Sample 2 should be assembled by raven including two internal polishing rounds,
followed by medaka and pilon polishing using the Illumina reads, and finally reference-based polishing with E. coli proteins using proovframe, which requires the file references-protein/Ecoli.faa
.
We therefore create the folders and files as follows:
.
├── assemblies
│ ├── sample1+filtlongMB500_flye+racon2+medaka+homopolish
│ ├── sample1_flye+racon2+medaka+homopolish
│ ├── sample2_raven2+medaka+pilon+proovframe
├── fastq-illumina
│ ├── sample2_R1.fastq
│ └── sample2_R2.fastq
├── fastq-ont
│ ├── sample1.fastq
│ └── sample2.fastq
├── references
│ └── Ecoli.fa
└── references-protein
└── Ecoli.faa
We also want to set the minimum read length threshold for Filtlong to 500nt and use the medaka model r941_min_high_g351
for both samples.
Therefore, we run the workflow with:
snakemake -s /opt/software/ont-assembly-snake/Snakefile --use-conda --cores 20 --config medaka_model=r941_min_high_g351 filtlong_min_read_length=500
Snakemake will recursively handle the dependencies for each assembly,
and create folders for all intermediate steps automatically.
Additionally, a symlink is created for each output assembly in the assemblies/
folder.
For the above example, the folders will look like this after running the workflow:
.
├── assemblies
│ ├── sample1+filtlongMB500_flye
│ ├── sample1+filtlongMB500_flye.fa -> sample1+filtlongMB500_flye/output.fa
│ ├── sample1+filtlongMB500_flye+racon2
│ ├── sample1+filtlongMB500_flye+racon2.fa -> sample1+filtlongMB500_flye+racon2/output.fa
│ ├── sample1+filtlongMB500_flye+racon2+medaka
│ ├── sample1+filtlongMB500_flye+racon2+medaka.fa -> sample1+filtlongMB500_flye+racon2+medaka/output.fa
│ ├── sample1+filtlongMB500_flye+racon2+medaka+homopolish
│ ├── sample1+filtlongMB500_flye+racon2+medaka+homopolishEcoli.fa -> sample1+filtlongMB500_flye+racon2+medaka+homopolish/output_Ecoli.fa
│ ├── sample1_flye
│ ├── sample1_flye.fa -> sample1_flye/output.fa
│ ├── sample1_flye+racon2
│ ├── sample1_flye+racon2.fa -> sample1_flye+racon2/output.fa
│ ├── sample1_flye+racon2+medaka
│ ├── sample1_flye+racon2+medaka.fa -> sample1_flye+racon2+medaka/output.fa
│ ├── sample1_flye+racon2+medaka+homopolish
│ ├── sample1_flye+racon2+medaka+homopolishEcoli.fa -> sample1_flye+racon2+medaka+homopolish/output_Ecoli.fa
│ ├── sample2_raven2
│ ├── sample2_raven2.fa -> sample2_raven2/output.fa
│ ├── sample2_raven2+medaka
│ ├── sample2_raven2+medaka.fa -> sample2_raven2+medaka/output.fa
│ ├── sample2_raven2+medaka+pilon
│ ├── sample2_raven2+medaka+pilon.fa -> sample2_raven2+medaka+pilon/output.fa
│ ├── sample2_raven2+medaka+pilon+proovframe
│ └── sample2_raven2+medaka+pilon+proovframeEcoli.fa -> sample2_raven2+medaka+pilon+proovframe/output_Ecoli.fa
├── fastq-illumina
│ ├── sample2_R1.fastq
│ └── sample2_R2.fastq
├── fastq-ont
│ ├── sample1.fastq
│ ├── sample1+filtlongMB500.fastq
│ └── sample2.fastq
├── references
│ └── Ecoli.fa
└── references-protein
└── Ecoli.faa
(Not shown is the content of each subfolder in assemblies/
and some auxiliary files.)
The assemblies generated by this workflow can be evaluated using the score-assemblies workflow.
One can either download the score-assemblies repository separately and just run it in the same folder after ont-assembly-snake is done, see the README or
add the Snakemake command line parameter --config run_score_assemblies=1
to the Snakemake call of ont-assembly-snake.