Compute SuSiE-inf TWAS weights

Extracts coefficients from an existing SuSiE-inf fit or fits `susieR::susie()` with `unmappable_effects = "inf"`.

Usage

susieInfWeights(
  X = NULL,
  y = NULL,
  susieInfFit = NULL,
  fitRetention = c("none", "slim", "full")
)

Arguments

X

Optional genotype matrix; when supplied it is only used to check that the fit covers the same number of variants.

y

Unused; retained for signature compatibility.

susieInfFit

Optional fitted SuSiE-inf object.

fitRetention

How much of the fit is kept on the returned weights: "none" attaches nothing, "slim" and "full" attach it as the "fit" attribute. These extractors hold no trimmable intermediates, so the two retaining levels behave alike here.

Value

Numeric vector of variant weights.

Non-zero weights with zero PIPs

SuSiE-inf decomposes effects into a mappable component (driven by `alpha * mu`, reported as per-variant PIPs) and an infinitesimal component (driven by `theta`). When the fit converges with no mappable effects – all `V` and `mu` zero, so every `pip == 0` – the returned weights are still non-zero because `susieR::coef.susie` adds `theta / X_column_scale_factors` to the mappable coefficient. This is intentional: it captures diffuse polygenic signal that the mappable component could not localize to any credible set. Consumers that interpret per-variant PIPs as a gate on whether to use the weights should be aware that low or zero PIPs do not imply zero TWAS weights here.

Examples

data(eqtlRegionExample)
X <- eqtlRegionExample$X[, 1:30]
y <- eqtlRegionExample$yRes
fit <- susieR::susie(X, y, L = 5)
#> HINT: nrow(X) = 415 >= 2 * ncol(X) = 60. Consider precomputing sufficient statistics with compute_suff_stat() and fitting with susie_ss() instead -- this avoids holding X in memory at every iteration and lets you reuse XtX across multiple y.
susieInfWeights(susieInfFit = fit)
#>  [1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0