Impute missing transcript counts using a shared nearest neighbor graph
Arguments
- combined_sce
A
SingleCellExperimentobject with gene counts and a "transcript" altExp slot.- dimred
The name of the reduced dimension to use for building the shared nearest neighbor graph.
- ...
Additional arguments to pass to
scran::buildSNNGraph. E.g.k = 30.
Details
For cells with NA values in the "transcript" altExp slot, this function imputes the
missing values from cells with non-missing values. A shared nearest neighbor graph is built using
reduced dimensions from the SingleCellExperiment object, and the imputation is done where
the imputed value for a cell is the weighted sum of the transcript counts of its neighbors.
Imputed values are stored in the "logcounts" assay of the "transcript" altExp slot.
The "counts" assay is used to obtain logcounts but left unchanged.
Examples
sce <- SingleCellExperiment::SingleCellExperiment(assays = list(counts = matrix(rpois(50, 5), ncol = 10)))
long_read <- SingleCellExperiment::SingleCellExperiment(assays = list(counts = matrix(rpois(40, 5), ncol = 10)))
SingleCellExperiment::altExp(sce, "transcript") <- long_read
SingleCellExperiment::counts(SingleCellExperiment::altExp(sce))[,1:2] <- NA
SingleCellExperiment::counts(SingleCellExperiment::altExp(sce))
#> [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
#> [1,] NA NA 4 4 9 7 6 7 5 1
#> [2,] NA NA 5 4 9 6 4 6 5 5
#> [3,] NA NA 5 5 9 3 7 8 4 2
#> [4,] NA NA 5 6 2 6 5 7 7 6
imputed_sce <- sc_impute_transcript(sce, k = 4)
#> Warning: more singular values/vectors requested than available
#> Warning: You're computing too large a percentage of total singular values, use a standard svd instead.
#> Imputing transcript counts ...
SingleCellExperiment::logcounts(SingleCellExperiment::altExp(imputed_sce))
#> 4 x 10 Matrix of class "dgeMatrix"
#> [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8]
#> [1,] 2.606167 2.387593 2.479993 2.479993 2.954196 2.985583 2.793234 2.686501
#> [2,] 2.686290 2.757838 2.749252 2.479993 2.954196 2.793234 2.308753 2.500984
#> [3,] 2.567505 2.568373 2.749252 2.749252 2.954196 1.987652 2.985583 2.850857
#> [4,] 2.656247 2.721574 2.749252 2.976074 1.321928 2.793234 2.571236 2.686501
#> [,9] [,10]
#> [1,] 2.627273 1.352516
#> [2,] 2.627273 3.132224
#> [3,] 2.362570 2.038135
#> [4,] 3.044394 3.367571