Closed jashapiro closed 2 years ago
An alternative solution may be to replace Matrix.utils::aggregate
with DelayedArray::rowsum()
which should have a similar function.
I think moving to fishpond
is still a good idea, but may not be required at this stage.
Describe the bug The removal of
Matrix.utils
from CRAN means we need to remove it as a dependency from our code, as the package will no longer buildThe only place that we use the function directly is in
collapse_intron_counts()
, which is in turn called byread_alevin()
andread_kallisto()
For
read_alevin()
, we should be able to use onfishpond
to perform the same functions for reading and collapsing directly.I don't think kallisto stores ambiguous counts separately, so we should be able to simply drop the intron reads and not use the
Matrix.utils::aggregate.Matrix()
function at all.We can probably perform this step within the
read_kallisto()
function itself and remove thecollapse_intron_counts()
function completely.