SNPLift takes a tab-delimited file, for example a VCF, with locus positions from a given genome and lifts over these positions so they match a new reference genome. The goal is to rapidly leverage the availability of new genomes without having to re-align all the sample reads and then call and filter the loci again.
In the process, a small proportion of the loci are inevitably lost. However, the transferred proportion is very high for genomes with low duplication content and when both genome versions are fairly similar. For example, our test run on real data gives a 99.87% transfers rate.
NOTE: Although SNPLift was designed primarily for VCFs containing SNP data,
any input file in which the two first columns contain chromosome names and
positions can be used. Lines beginning with a hash sign (#) are simply
transferred to the output file without modification and any column beyond the
first two are also left unchanged (there is an option to update SNP IDs in the
config file). As such, SNPLift will work with any marker type or even bed
files, as long as the two first columns contain chromosome and position
information and that there are other columns with informations to transfer.
WARNING: For local regions that differ between the two assemblies, a small proportion of the SNPs (from 0.5 to 1% in our tests) will end up with an approximate position. Further work will be done to improve this, but ultimately the only way to guaranty that all the positions on the new genome are correct is to re-align the reads and call the genotypes again. Note that positions within one window length of the end of contigs will also get an erroneous new position. As such, transfering positions to a lot of small contigs could lead to a high proportion of errors in positions on the new reference.
See licence information at the end of this file.
If you use SNPLift, please cite the following paper:
https://www.biorxiv.org/content/10.1101/2023.06.13.544861v1
When transferring millions of positions, once the genome is indexed, SNPLift will typically transfer between 0.5 and 1 million positions per minute. For example, 50M Zea mays SNPs were transferred in 53m53s, or about 0.98 million SNPs transfered per minute. For datasets with less than 1M positions, using 1 to 10 CPUs is recommended. Above that, run time will decrease up to 40 CPUs, but the overall efficiency is also reduced. Overall, using 10 CPUs is always a good choice and values above 20 CPUs will be more wasteful of ressources, even on large datasets. See article in the Citation section above for more benchmark details.
SNP datasets and genomes used to benchmark SNPLift program can be found on Dryad: https://doi.org/10.5061/dryad.h9w0vt4nx
To use SNPLift, you will need a local copy of its repository. Different
releases can be found here
under the Tags tab. It is recommended to always use the latest release or
even the development version. You can either download an archive of the latest
release at the above link or get the latest commit (recommended) with the
following git command:
git clone https://github.com/enormandeau/snplift
To run SNPLift, you will need to have the following programs installed. To facilitate installation, an effort has made to use only easy-to-install programs, as well as avoid non standard libraries for Python and R code. Only scipy is required for Python.
- SNPLift will only work on GNU Linux or OSX
- python 3.5+ and scipy
- R 3+ (ubuntu/mint:
sudo apt-get install r-base-core) - bash 4+
- gnu parallel
- bwa
- samtools
- git (to clone this repository and the test dataset)
- minimap2 (to visualize the collinearity of the two genomes)
- epstopdf (to visualize the collinearity of the two genomes)
- miniasm (using minidot to visualize the collinearity of the two genomes)
NOTE: git is required for the test run since it is used to get the test
data.
Before trying SNPLift on your data, we suggest running it on the prepared test dataset. This will confirm that you have all the required dependencies.
The test dataset consists in the first chromosome from two different genome assemblies from Medicago truncatula and a VCF with SNPs found in the first chromosome of the reference genome. The VCF contains the genotypes of 10 samples for 190,443 SNPs. The test, which includes downloading the data and indexing the genome, takes about 53s on 10 Xeon processors from 2016.
You can run the full SNPLift test with:
./01_scripts/util/run_test.sh
- Install dependencies
- Download a copy of the SNPLift repository (see Installation above)
- Make sure that the chromosome names used in the file (eg: VCF) match EXACTLY
those found in the
old_genome.fastafile before any space character. - Modify
02_infos/snplift_config.sh
During the analyses, the following steps are performed in this order:
- Optional: Visualize collinearity of the two genomes (slow, using minimap)
- Index new genome (can be skipped if already indexed)
- Get original coordinates
- Extract flanking sequences around SNPs from the old genome
- Map reads with bwa to new genome
- Extract features from alignments
- Optional (for debugging purposed): Visualize alignment features
- Filters based on alignment features
- Save some bad alignments if they are locally collinear with good ones
- Update coordinates
- Update input file (eg: VCF)
When using multiple CPUs, some of these steps are run on parallel.
Once SNPLift is installed and the dependencies are met, you need to do the following:
- Prepare your input files
- Modify the config file
- Launch SNPLift with:
time ./snplift 02_infos/snplift_config.sh
The output of SNPLift is a file (eg: VCF) in which the positions for which a good alignment was found are transferred to the coordinates of a new reference genome.
Optionally, if CHECK_COLLINEARITY is set to 1, a dot plot collinearity
figure in .eps and .pdf formats is produced.
Optionally, if SKIP_VISUALIZATION is set to 0, a figure showing some of the
features used for filtering the alignments is produced. This is used for
debugging purposes.
Some additinal parameter in the config file permit to do final corrections to VCF files.
SNPLift works well when the two genome versions are more similar differences. As the differences between the orthologous sequences increase, the proportion of SNPs that can be transferred decreases. Whole or partial genome duplication will also have an impact on the capacity to transfer SNPs between assemblies.
For SNPs with position within 300bp of scaffold ends (or the value of WINDOW_LENGTH in the configuration file), the reported position in the new file (eg: VCF) can be slightly off.
In our tests, about 1% of the transfered positions were not exact (see reference in the Citation section).
CC share-alike
SNPLift by Eric Normandeau is licensed under a
Creative Commons Attribution-ShareAlike 4.0 International License.
