Wrapper function for MAGIC imputation parameters
params_sc_magic.RdParameters controlling the MAGIC imputation run by
run_magic_sc(). The defaults mirror the reference implementation.
Usage
params_sc_magic(
n_steps = 3L,
clip_threshold = 0.01,
gene_batch_size = 1000L,
layer = c("norm", "raw"),
allow_large = FALSE
)Arguments
- n_steps
Integer. Diffusion steps applied to the counts. Zero is legal and hands back the un-imputed values, which is a cheap way to compare the two. Defaults to
3L.- clip_threshold
Numeric. Imputed values below this are zeroed after the last step. Defaults to
0.01.- gene_batch_size
Integer. Genes streamed off the binary store per block. Bounds the scratch memory and is clamped to the number of requested genes. Defaults to
1000L.- layer
String. Which stored layer to impute. The operator preserves per-cell mass, so imputed values sit on the scale of whatever went in: imputing raw counts and imputing log-normalised counts are different operations rather than the same one rescaled. One of
c("norm", "raw"). Defaults to"norm".- allow_large
Boolean. Skip the output size guard. The dense output is capped at 1e9 elements, i.e. 4 GB of
f32. Defaults toFALSE.
Value
A named list with the following elements:
n_steps - Integer. Diffusion steps applied to the counts. Zero is legal and hands back the un-imputed values, which is a cheap way to compare the two. Defaults to
3L.clip_threshold - Numeric. Imputed values below this are zeroed after the last step. Defaults to
0.01.gene_batch_size - Integer. Genes streamed off the binary store per block. Bounds the scratch memory and is clamped to the number of requested genes. Defaults to
1000L.layer - String. Which stored layer to impute. The operator preserves per-cell mass, so imputed values sit on the scale of whatever went in: imputing raw counts and imputing log-normalised counts are different operations rather than the same one rescaled. One of
c("norm", "raw"). Defaults to"norm".allow_large - Boolean. Skip the output size guard. The dense output is capped at 1e9 elements, i.e. 4 GB of
f32. Defaults toFALSE.