Impute missing gene statistics across studies

Handles genes not measured in all studies by applying one of four imputation strategies before meta-analysis.

Usage

mx_impute(de_results, method = c("exclude", "mean", "knn", "weighted"))

Arguments

de_results

A named list of data.frame objects from mx_de_all.

method

Character scalar. Imputation strategy. One of:

"exclude": Exclude genes not present in all studies (most conservative; default).
"mean": Impute missing log2FC with the mean across available studies; impute p-values as 1.
"knn": K-nearest-neighbour imputation of log2FC across genes with similar cross-study fold-change profiles.
"weighted": Mean log2FC weighted by gene coverage.

Value

The input list with missing values filled according to the chosen strategy.

Caution

The "mean", "knn" and "weighted" strategies fill a study's missing fold change with values derived from the other studies. This is circular: it inflates the apparent number of studies per gene, shrinks heterogeneity (I-squared) and has no standard error, so imputed rows are ignored by effect-size methods. For inference, prefer mx_meta with min_studies, which pools each gene over the studies that measured it, or mx_filter_coverage. Imputation is mainly useful for visualising complete matrices.

References

Villatoro-García, J.A. et al. (2022) DExMA: an R package for performing gene expression meta-analysis with missing genes. Mathematics, 10(18), 3376. doi:10.3390/math10183376

Examples

if (FALSE) { # \dontrun{
  de_results <- mx_impute(de_results, method = "mean")
} # }