Skip to contents

Wrapper function for parameters for the limma-voom workflow

params_limma_voom.Rd

Parameters for the limma linear model chain, implemented in Rust via the edge-rs crate and gated against limma 3.66.0. Defaults are limma's own, except for filter: inside BulkDge() the genes were already filtered by qc_bulk_dge(). route = "voom" is voomLmFit(): precision weights from the mean-variance trend, then weighted least squares. route = "trend" is limma-trend: log-CPM straight into lmFit(), with the trend absorbed by eBayes(trend = TRUE). The empirical Bayes trend follows the route.

Usage

params_limma_voom(
  route = c("voom", "trend"),
  norm_method = c("TMM", "TMMwsp", "RLE", "upperquartile", "none"),
  filter = FALSE,
  min_mean = 0,
  robust = FALSE,
  prior_count = NULL,
  adaptive_span = TRUE,
  span = 0.5,
  proportion = 0.01
)

Arguments

route

String. Whether to run the voom or the limma-trend route. One of c("voom", "trend"). Defaults to "voom".

norm_method

String. Library size normalisation. One of c("TMM", "TMMwsp", "RLE", "upperquartile", "none"). Defaults to "TMM".

filter

Boolean. Run filterByExpr() before fitting. Defaults to FALSE.

min_mean

Numeric. Drop genes whose mean count across samples is below this. Applied on top of filter. Defaults to 0.0.

robust

Boolean. Robust empirical Bayes, eBayes(robust = TRUE). Defaults to FALSE.

prior_count

Numeric or NULL. Count added before the log. NULL takes the route's own default, 0.5 for voom and 2 for trend. Defaults to NULL.

adaptive_span

Boolean. Derive the lowess span from the number of genes, as limma does since 3.56. Only used by voom. Defaults to TRUE.

span

Numeric. Lowess span for the voom trend, only read if adaptive_span = FALSE. Defaults to 0.5.

proportion

Numeric. Assumed proportion of differentially expressed genes, only used for the B-statistic. Defaults to 0.01.

Value

A named list with the following elements:

  • route - String. Whether to run the voom or the limma-trend route. One of c("voom", "trend"). Defaults to "voom".

  • norm_method - String. Library size normalisation. One of c("TMM", "TMMwsp", "RLE", "upperquartile", "none"). Defaults to "TMM".

  • filter - Boolean. Run filterByExpr() before fitting. Defaults to FALSE.

  • min_mean - Numeric. Drop genes whose mean count across samples is below this. Applied on top of filter. Defaults to 0.0.

  • robust - Boolean. Robust empirical Bayes, eBayes(robust = TRUE). Defaults to FALSE.

  • prior_count - Numeric or NULL. Count added before the log. NULL takes the route's own default, 0.5 for voom and 2 for trend. Defaults to NULL.

  • adaptive_span - Boolean. Derive the lowess span from the number of genes, as limma does since 3.56. Only used by voom. Defaults to TRUE.

  • span - Numeric. Lowess span for the voom trend, only read if adaptive_span = FALSE. Defaults to 0.5.

  • proportion - Numeric. Assumed proportion of differentially expressed genes, only used for the B-statistic. Defaults to 0.01.

References

Law, et al., Genome Biol, 2014; Smyth, Stat Appl Genet Mol Biol, 2004