Introduction to prewas
prewas is a tool to standardize the pre-processing of your genomic data before performing a bacterial genome-wide association study (bGWAS).
prewas creates a variant matrix (where each row is a variant, each column is a sample, and the entries are presence - 1 - or absence - 0 - of the variant) that can be used as input for bGWAS tools, such as hogwash or treeWAS.
When creating the binary variant matrix, prewas can perform 3 pre-processing steps including:
dealing with multiallelic SNPs,
(optional) dealing with SNPs in overlapping genes, and
choosing a reference allele.
These 3 steps in the prewas workflow are shown below. (B) shows how prewas handles multiallelic sites, (C) shows options for choosing a reference allele and (D) shows how prewas can group SNPs into genes. A detailed workflow indicating required file inputs can be seen below in the “Detailed prewas workflow” section.
prewas workflow
1. Dealing with multiallelic SNPs
A multiallelic site is a site with more than two alleles present at a locus. Most bGWAS methods require a binary input (variant presence or absence) - so multiallelic sites don’t fit into this mold.
One could consider the reference allele 0 and any alternative allele as 1. However, each alternative allele could confer a different functional impact on the expressed protein. Thus, lumping the alternative alleles into one category will take away the ability to study these functional differences.
prewas uses the multi-line format to represent mutliallelic SNPs. Below are examples of how a triallelic site (T, A, and G) can be represented in a single line of a variant matrix compared to multiple lines of a variant matrix.
Single-line format:
- T -> A, G
Multi-line format:
T->A
T->G
2. Dealing with SNPs in overlapping genes
A SNP in an overlapping gene could have a different impact on those two genes. Therefore, when a gff is provided, prewas will output a binary variant matrix where a SNP in x overlapping genes will be represented on x lines. When a gff is provided, prewas will also output a gene-based variant matrix indicating presence or absence of at least 1 variant in that gene, and SNPs in overlapping genes will be assigned to both genes.
3. Choosing a reference allele
A reference allele, the allele that will be denoted 0 in a binary matrix could be:
- the allele in a reference genome
- the major allele at each position
- the ancestral allele at each position
Choosing a reference allele can be particularly important when doing gene-based analysis and therefore aggregating SNPs by gene. We suggest referencing to the ancestral allele for evolutionary interpretability. In cases where ancestral reconstruction is not feasible (e.g. computational intensity) or low confidence, we suggest referencing to the major allele instead of referencing to an arbitrary reference genome because using the major allele leads to less masking of variation at the gene level.
Reference to the major allele to avoid masking variation at the gene level
When ancestral reconstruction is not used (anc = FALSE), prewas will use the major allele as the reference allele.
