Homopolish is a genome polisher originally developed for Nanopore and subsequently extended for PacBio CLR. It generates a high-quality genome (>Q50) for virus, bacteria, and fungus. Nanopore/PacBio systematic errors are corrected by retreiving homologs from closely-related genomes and polished by an SVM. When paired with Racon and Medaka, the genome quality can reach Q50-90 (>99.999%) on Nanopore R9.4/10.3 flowcells (Guppy >3.4). For PacBio CLR, Homopolish also improves the majority of Flye-assembled genomes to Q90 (see Accuracy).
- Installation
- Quick usage for polishing ONT homopolymer errors
- Quick usage for polishing ONT modification-mediated errors
- Accuracy
- Citation
- Contact
Homopolish is recommendated to install and run within a conda environment
git clone https://github.com/ythuang0522/homopolish.git
cd homopolish
conda env create -f environment.yml
conda activate homopolish
Homopolish is available in bioconda(flexible mode only).
conda config --set channel_priority flexible
conda create -n homopolish -c conda-forge -c bioconda homopolish=0.4.1=pyhdfd78af_1
If conda install takes too long, you can try mamba for installation.
mamba create -n homopolish -c conda-forge -c bioconda homopolish
Homopolish retrieves homologous sequences by scanning microbial genomes compressed in (Mash) sketches. Three sketches, bacteria (3.3Gb, updated on 2022/8/19) , virus (74Mb), and fungi (74Mb) can be downloaded from the following URLs using wget or curl. The previous smaller size of bacterial sketch (720Mb) is available here.
Bacteria: http://140.123.104.107/bioinfo/mash_sketches/bacteria.msh.gz
Virus: http://140.123.104.107/bioinfo/mash_sketches/virus.msh.gz
Fungi: http://140.123.104.107/bioinfo/mash_sketches/fungi.msh.gz
Then unzip the downloaded skeches.
gunzip bacteria.msh.gz
Homopolish should be run with a pre-trained model (R9.4.pkl/R10.3.pkl for Nanopore and pb.pkl for PacBio CLR) and one sketch (virus, bacteria, or fungi). For Nanopore sequencing, Homopolish should be run after the Racon-Medaka pipeline as it only removes indel errors. For PacBio CLR sequencing, it can be invoked directly after Flye assembly. For instance, if your Medaka-polished genome (yourgenome.fasta) is bacteria and sequenced by R9.4 flowcell, please type
python3 homopolish.py polish -a yourgenome.fasta -s bacteria.msh -m R9.4.pkl -o youroutput
If installed from bioconda, you should directly invoke the script.
homopolish polish -a yourgenome.fasta -s bacteria.msh -m R9.4.pkl -o youroutput
You can specify particular genus and species via -g (Taxanomic names in NCBI), which will randomly select genomes of the same genus/species for polishing. However, we don't recommend using this argument in most cases as the autosearch often identifies more closely-related strains.
python3 homopolish.py polish -a yourgenome.fasta -g genusname_speciesname -m R9.4.pkl -o youroutput
If you wanna use private local genomes instead of NCBI, specify the path to your own (fasta) database via -l.
python3 homopolish.py polish -a yourgenome.fasta -l path_to_your_genomes.fasta -m R9.4.pkl -o youroutput
Run python3 homopolish.py polish --help to view all the options:
usage: homopolish.py polish [-h] -m MODEL_PATH -a ASSEMBLY
(-s SKETCH_PATH | -g GENUS) [-t THREADS]
[-o OUTPUT_DIR] [--minimap_args MINIMAP_ARGS]
[--mash_threshold MASH_THRESHOLD]
[--download_contig_nums DOWNLOAD_CONTIG_NUMS] [-d]
[--mash_screen]
(-l LOCAL_DB_PATH)
optional arguments:
-h, --help show this help message and exit
-m MODEL_PATH, --model_path MODEL_PATH
[REQUIRED] Path to a trained model (pkl file). Please
see our github page to see options.
-a ASSEMBLY, --assembly ASSEMBLY
[REQUIRED] Path to a assembly genome.
-s SKETCH_PATH, --sketch_path SKETCH_PATH
Path to a mash sketch file.
-g GENUS, --genus GENUS
Genus name
-l LOCAL_DB_PATH, --local_DB_path LOCAL_DB_PATH
Path to your local DB (ex: cat closely-related_genomes1.fasta closely-related_genomes2.fasta> DB.fasta)
-t THREADS, --threads THREADS
Number of threads to use. [1]
-o OUTPUT_DIR, --output_dir OUTPUT_DIR
Path to the output directory. [output]
--minimap_args MINIMAP_ARGS
Minimap2 -x argument. [asm5]
--mash_threshold MASH_THRESHOLD
Mash output threshold. [0.95]
--ani Ani identity [99%]
--download_contig_nums DOWNLOAD_CONTIG_NUMS
How much contig to download from NCBI. [20]
-d, --debug Keep the information of every contig after mash, such
as homologous sequences and its identity infomation.
[no]
--mash_screen Use mash screen. [mash dist]
The program's output files will be saved in the folder named youroutput (-o youroutput ), and the polished genoem is named yourgenome_homopolished.fasta.
Homopolish impleneted a submodule called modpolish for correcting modification-mediated errors. Given a draft genome with ONT reads (fastq), modpolish correcting the modification-mediated errors using reads, quality and homologs.
python3 homopolish.py modpolish -a yourgenome.fasta -q fastq_path -m R9.4.pkl -s bacteria.msh -o youroutput
You can supply the bam (i.e., reads to draft genome) insetad of reads for skipping the time-consuming alignment.
python3 homopolish.py modpolish -a yourgenome.fasta -b bam_path -m R9.4.pkl -s bacteria.msh -o youroutput
Run python3 homopolish.py modpolish --help to view all the options:
usage: python3 homopolish.py modpolish -a ASSEMBLY -s SKETCH_PATH
[ -q FASTQ | -b BAM ]
(-o OUTPUT_DIR)
(-p SPECAIL_PATTERN)
(-t THREADS)
(-l LOCAL_DB_PATH)
optional arguments:
-h, --help show this help message and exit
-a ASSEMBLY, --Draft genome file path
[REQUIRED] Path to a assembly genome.
-s SKETCH_PATH, --sketch_path SKETCH_PATH
Path to a mash sketch file.
Genus name
Path to your local DB (ex: cat closely-related_genomes1.fasta closely-related_genomes2.fasta> DB.fasta)
-o OUTPUT_DIR, --output_dir OUTPUT_DIR
Path to the output directory. [output]
-q FASTQ, fastq File
-b BAM, --bam Path to a bam file
-t THREADS, --threads THREADS
Number of threads to use. [1]
-p SPECAIL_PATTERN, string of supported motif: CCAGC or GCAGC ...
-l LOCAL_DB_PATH, --local_DB_path LOCAL_DB_PATH
Path to your local DB (ex: cat closely-related_genomes1.fasta closely-related_genomes2.fasta> DB.fasta)
A genome named yourgenome_modpolished.fasta will be outputted.
If you use the parameter -d, directory content in a tree-like format is below.
homologous_sequencescontains other homologous speciesAll_homologous_sequences.fna.gzwhich concatenate all file inhomologous_sequences
├── yourgenome_homopolished.fasta
└── debug
├── contig_1_segment0
│ ├── All_homologous_sequences.fna.gz
│ ├── contig_1_segment0.fasta
│ ├── contig_1_segment0.feather
│ ├── contig_1_segment0.npz
│ ├── contig_1_segment0.paf
│ ├── contig_1_segment0.sort.tab
│ ├── homologous_sequences
│ │ ├── GCF_000775955.1_ASM77595v1_genomic.fna.gz
│ │ └── .......
│ ├── polished.fasta
│ └── result.feather
├── contig_2_segment0
│ └── ......
└── ......
Comparison of genome accuracy polished by Racon, Medaka, MarginPolish, HELEN, and Homopolish on Nanopore R9.4. Median Q scores were computed by fastmer. We note that these are based on early ONT basecaller (Guppy 3.2). After Guppy 3.4, we have seen consistent >Q50 genomes.
Since v0.3, we found FastANI is more accurate than mash for selecting highly-similar genomes for polishing. The FastANI version improved the accuracy of R9.4 and R10.3 (see below Guppy 3.2 results). We note that short contigs/plasmids (<~6kb) won't benefit from FastANI as it can't output distance estimatino.
With the addition of FastANI, genomes sequenced by PacBio CLR (and assembled by Flye) can now be also siginificantly improved by Homopolish.
If you use homopolish, please cite
Huang, Y.-T., Liu, P.-Y., and Shih, P.-W. Homopolish: a method for the revmoal of systematic errors in nanopore sequencing by homologous polishing, Genome Biology, 2021.
ythuang at cs.ccu.edu.tw