We expect a feature_selection
HDF5 group at the root of the file, containing statistics for highly variable gene detection from the RNA data.
The group itself contains the parameters
and results
subgroups.
In multi-modal contexts, the results here are only relevant for the RNA modality.
Definitions:
num_genes
: number of genes in the RNA expression matrix.
parameters
should contain:
span
: a scalar float specifying the span to use for the LOWESS smoother. This should lie between 0 and 1.
results
should contain:
means
: a 1-dimensional float dataset of length equal tonum_genes
, containing the mean log-expression of each gene.vars
: a 1-dimensional float dataset of length equal tonum_genes
, containing the variance in log-expression of each gene.fitted
: a 1-dimensional float dataset of length equal tonum_genes
, containing the fitted value of the trend for each gene.resids
: a 1-dimensional float dataset of length equal tonum_genes
, containing the residuals from the trend for each gene.
Added in version 1.0.