Make some standardized tables from gwas and annotation tables

Usage

make_panvar_tables(
  gwas.res,
  qtl.df = NULL,
  tag.snp = NULL,
  annotation.table,
  plink.path = NULL,
  pvals.in.log = T,
  geno.bed.filename,
  geno.bed.directory = "/.",
  temp.dir = tempdir(),
  window,
  snp.to.gene.vars = "LD",
  snp.to.gene.buffer = 0,
  compute.scores = F,
  score.vars = NULL,
  score.dirs = NULL,
  score.weights = NULL
)

Arguments

gwas.res: data.frame of all gwas results, should contain columns (CHR, POS, PVAL), corresponding to (chromosome, physical position, and pvalue).
qtl.df: data.frame, table that includes list of snps to calculate LD to with columns (CHR, POS, LOGPVAL), corresponding to (chromosome, physical position, and -log10(p-value)). QTL are typically defined as hits grouped by LD by something like plink --clump. See get_ld_in_window
tag.snp: character, marker.ID of snp around which to calculate LD. In the form 'CHR-POS'
annotation.table: table with annotations with columns (geneID, CHR, start, end, annotation). start and end correspond to base-pair coordinates of start and end of gene. CHR is chromosome of gene.
plink.path: character, optional, path to plink2 executable. Will overide option set by set_plink_path.
pvals.in.log: boolean, if TRUE PVAL column has already been converted to -log10(pvalue)
geno.bed.filename: character, prefix of genotype files in plink (bed/bim/fam) format. Do not include ".bed" extension.
geno.bed.directory: character, directory where genotype files are located
temp.dir: character, where to output some temporary files.
window: numeric, total window size in KB, all variants within .5 * window are calculated.
snp.to.gene.vars: character, numeric variables in gwas.res to aggregate by gene. For each gene, snps with a physical position with the start and end of the gene are considered. The maximum value for all snps within the gene is returned. Special values, DIST, LD and LOGPVAL can be included in addition to any user supplied variables.
snp.to.gene.buffer: numeric, kilobases to add to gene start and end to include genes that are close but not in gene. Snpeff uses 5 KB by default to call a snp "upstream"/"downstream" variant. default is 0.
compute.scores: boolean, if TRUE, snp scores will be computed. See details for more info.
score.vars: character, vector of column names indicating which variables to included in the score. If compute.scores is TRUE and score.vars is NULL, the default score will use equally weighted variables: "DIST", "LOGPVAL", "LD".
score.dirs: numeric, a vector indicating which direction is to be considered more indicative of an association. 1 indicates higher is better, -1 indicates lower is better. The order should correspond with the order in cols.
score.weights: numeric, a vector indicating weights for the variables. These must add up to 1.

Value

A named list with the following entries:

gwas: formatted gwas results.
anno: formatted annotation results.
key.snp: tag.snp or in the qtl.df case the highest p-value snp supplied to the function for downstream use.
qtl.df: qtl.df supplied to the program if used.

Details

Scores: Scores are simple scaled and weighted averages of some variables. First variables are normalized using min/max normalization. The variables are then made negative if they need to be reversed to indicate a larger value as a more desirable value.

For example, distance from the key snp should be reversed as a small distance is more desirable. A log-pvalue is already of this form 'bigger is better' so does not need to be altered.

Finally, a weighted average is taken based on user defined weights. The default weights all variables equally. The outcome is a score from 0-1 that ranks the snps based on these variables. see: make_scores

Examples

# work in progress