Skip to content

Code for building and testing variant ranking strategies

License

Notifications You must be signed in to change notification settings

langmead-lab/FORGe

Repository files navigation

FORGe tool for ranking variants and building an optimal graph genome

FORGe consists of 2 primary scripts -- rank.py and build.py -- as well as a helper script vcf_to_1ksnp.py for generating input files in the proper format.

vcf_to_1ksnp.py

FORGe takes information about genetic variants in a 1ksnp and an optional phasing file. The 1ksnp format is less complex than VCF, but you can convert VCF to using the included vcf_to_1ksnp.py.

In a 1ksnp file, each alternate allele is stored as a separate line with the following columns:

Chromosome	Position	Reference Allele	Alternate Allele	Population Frequency	??	# Alternates	Variant Name

The phasing file is a comma-separated table with a column for each individual and row for each variant, in the same order they appear in the 1ksnp file. An element contains the allele for the corresponding individual (column) and variant (row), with 0 indicating the reference allele and k indicating the kth alternate allele, in order, appearing in the 1ksnp file.

FORGe includes a helper script vcf_to_1ksnp.py to facilitate convertsion from a VCF file and set of ingroup individuals to FORGe variant and phasing files. Ingroup individuals can be specified as a list of names to either include (--ingroup) or exclude (--outgroup).

Example usage:

./vcf_to_1ksnp.py --reference ref.fa --vcf variants.vcf --ingroup names.txt --out variants.1ksnp --individuals phasing.txt

rank.py

rank.py takes a linear reference genome fasta, a 1ksnp variant file and an optional file containing phasing information. The user also specifies a model using --method (options are: popcov, popcov-blowup, or hybrid) and a window size using --window-size. Note that the hybrid model produces scores both with and without "blowup avoidance," whereas popcov and popcov-blowup do this separately.

The user can indicate a specific chromosome to process using --chrom. Otherwise the full genome is used. When running the hybrid ranking method for window sizes over ~35, we recommend adding the --prune which limits blowup in regions dense with genetic variants. A limit of 15 variants per window performed well in practice.

Example usage:

./rank.py --method popcov --reference ref.fa --vars variants.1ksnp --window-size 100 --phasing phasing.txt --output ordered.txt

build.py

build.py takes as input a set of ranked variants (output by rank.py) and a percentage of variants to include in the graph. It produces the necessary input files to build an index with HISAT2 or with Bowtie (ERG).

Example usage:

./build.py --reference ref.fa --vars variants.1ksnp --window-size 100 --hisat variants.snp --sorted ordered.txt --pct 10

Full pipeline

From beginning to end, running the FORGe pipeline with HISAT2 might look like this:

./vcf_to_1ksnp.py --reference ref.fa --vcf variants.vcf --ingroup names.txt --out variants.1ksnp --individuals phasing.txt
./rank.py --method popcov --reference ref.fa --vars variants.1ksnp --window-size 100 --phasing phasing.txt --output ordered.txt
./build.py --reference ref.fa --vars variants.1ksnp --window-size 100 --hisat variants.snp --sorted ordered.txt --pct 10
$HISAT_HOME/hisat2-build --snp variants.snp ref.fa index