Reference flow is a computational genomics pipeline to align short sequencing reads with a number of reference genomes built with population genomic information included.
The preprint "Reducing reference bias using multiple population reference genomes" is available on bioRxiv.
The experiments we've done are provided in the Reference Flow Experiments repository.
Reference flow is written using Snakemake and it can be installed using conda:
conda install -c bioconda -c conda-forge snakemake
Other installation approaches are also provided in the Snakemake installation page.
Population information for the 1000 Genomes Project individuals can be downloaded by running the script:
sh src/download_1kg_pop_table.sh
LevioSAM is used to perform coordinate system mappign between population reference genomes and a standard reference genome such as GRCh38. The levioSAM software tracks the "edits" between a pair of genomes using a VCF file. Please refer to the levioSAM github page to install the software.
We hosted pre-built RandFlow-LD (22GB) and RandFlow-LD-26 (96GB) indexes.
Script src/download_prebuilt_indexes.sh helps to download the RandFlow-LD indexes, including an indexed major-allele reference and 5 indexed superpop references.
sh src/download_prebuilt_indexes.sh
# Run `sh src/download_prebuilt_indexes.sh randflow_ld_26` to download pre-built RandFlow-LD-26 indexes.
cd snakemake
snakemake -j 32
Before executing the Snakemake pipeline, it is recommended to perform a dry-run snakemake -np to preview the scheduling plan.
The option -j specifies the number of threads used, which is set to 32 in the above example.
By default, a directory called run will be created under snakemake and all the results will be under it.
The alignment results are aggrelated as a single SAM file, which uses the GRCh38 coordinate system by default.
The final SAM file will appear in directory snakemake/run/experiments/test/thrds0_S1_b1000_ld1/wg-refflow-10-thrds0_S1_b1000_ld1-liftover.sam.
To use another read set, please change the READS1 parameter in snakemake/config.yaml,
or run command:
snakemake -j 32 --config READS1=<file>
By set the USE_PREBUILT option to False, users can run the complete reference flow pipeline.
The GRCh38 reference genome can be downloaded from NCBI using the following script:
sh src/download_genome.sh
Users may choose other reference genomes of interest by changing the GENOME parameter in snakemake/config.yaml.
The 1000 Genomes Project GRCh38 call sets with SNVs and indels can be downloaded using script:
sh src/download_1kg_vcf.sh
Users may switch to the 1KG phase-3 call set by using the corresponding VCFs and changing parameters DIR_VCF, VCF_PREFIX, and VCF_SUFFIX in the configuration file.
For now we only tested on call sets provided by the 1000 Genomes Project.
Call sets provided by other studies may differ slightly in population labelling and VCF format, causing the pipeline not working properly.
Parameters for reference flow are specified in snakemake/config.yaml and described here.
We recommend users try a dry-run by snakemake -np to check if the pipeline is ready.
When all set, run
snakemake -j 32