Skip to contents

Standardize and QC GWAS summary statistics

Source: R/instruments_exposure.R
qc_gwas_sumstats.Rd

Renames common GWAS summary-statistic column conventions to the bigsnpr convention (chr, pos, a0, a1, beta, beta_se, n_eff, freq) and applies the standard LDpred2 QC filters: missing/invalid values, extreme effect sizes, ambiguous-strand (A/T and C/G) SNPs, and — when effect allele frequencies are available — the standard-deviation ratio check from the LDpred2 tutorial.

Usage

qc_gwas_sumstats(
  sumstats,
  n_eff = NULL,
  drop_ambiguous = TRUE,
  beta_iqr_mult = 10,
  sd_ratio_range = c(0.5, 2),
  sd_y = NULL
)

Arguments

sumstats

A data.frame of GWAS summary statistics. Recognized column aliases: chr/chromosome/CHR; pos/base_pair_location/ BP; a1/effect_allele/ALT; a0/other_allele/REF; beta/BETA/effect/logOR; beta_se/standard_error/se/SE; freq/effect_allele_frequency/EAF/FRQ; n_eff/n/N; p/p_value/P; rsid/rs_id/SNP/ID.

n_eff

Optional scalar: effective sample size, used when the input has no sample-size column.

drop_ambiguous

Logical: drop A/T and C/G SNPs whose strand cannot be resolved. Default TRUE.

beta_iqr_mult

Numeric: drop SNPs with \(|\beta - \mathrm{median}(\beta)| >\) beta_iqr_mult \(\times \mathrm{IQR}(\beta)\). Default 10. Set to Inf to skip.

sd_ratio_range

Numeric length-2: acceptable range for the ratio of the summary-statistic-implied genotype SD to the frequency-implied SD, \(\sqrt{2 f (1-f)}\). SNPs outside the range are dropped. Default c(0.5, 2) (LDpred2 tutorial). Requires freq; skipped with a message when frequencies are absent.

sd_y

Optional scalar: phenotypic SD used in the SD-ratio check. When NULL (default), estimated as the 1st percentile of \(\sqrt{0.5 (n_{eff} \cdot se^2 + \beta^2)}\) across SNPs, the continuous-trait approximation used in the LDpred2 tutorial.

Value

A list with:

sumstats

data.frame of QC-passing variants with standardized columns chr, pos, a0, a1, beta, beta_se, and (when available) n_eff, freq, p, rsid.

qc

data.frame with counts of input variants and variants dropped by each filter.

sd_ratio

numeric vector of SD ratios for retained variants, or NULL when the check was skipped.

Examples

ss <- data.frame(
  chromosome = 1, base_pair_location = 1000 + 0:9 * 1000L,
  effect_allele = rep(c("A", "C"), 5), other_allele = rep(c("G", "T"), 5),
  beta = rnorm(10, 0, 0.05), standard_error = 0.01,
  effect_allele_frequency = runif(10, 0.1, 0.5)
)
out <- qc_gwas_sumstats(ss, n_eff = 50000)
str(out$sumstats)
#> 'data.frame':	10 obs. of  8 variables:
#>  $ chr    : chr  "1" "1" "1" "1" ...
#>  $ pos    : int  1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
#>  $ a1     : chr  "A" "C" "A" "C" ...
#>  $ a0     : chr  "G" "T" "G" "T" ...
#>  $ beta   : num  0.0253 0.0591 -0.0423 -0.0497 -0.0457 ...
#>  $ beta_se: num  0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01
#>  $ freq   : num  0.211 0.429 0.246 0.388 0.185 ...
#>  $ n_eff  : num  50000 50000 50000 50000 50000 50000 50000 50000 50000 50000
out$qc
#>   n_input dropped_missing dropped_extreme_beta dropped_ambiguous
#> 1      10               0                    0                 0
#>   dropped_sd_ratio n_output
#> 1                0       10