Adding a custom DimReduction for the projection of cells into a pre-existing low dimensional space

[RegnerM2015]

Hi @rcorces and @jgranja24

I am trying to figure out if I can replicate the approach taken in this project (https://github.com/GreenleafLab/MPAL-Single-Cell-2019). In this work, a normal reference is created using only healthy cells. Then, disease cells are projected into the same low-dimensional space using the variable features of the normal reference. This framework is very different from just clustering/processing healthy and disease cells together as this explores a different and wider range of variable features.

Is it possible to implement this framework in the ArchR workflow? I think it would require a function to add a custom DimReduction similar to the one mentioned here: #572

I found the .projectLSI function from the source code which could be used to perform such a task.

ArchR/R/IterativeLSI.R

Line 1234 in 968e442

.projectLSI <- function(

Thank you!

[rcorces]

I agree that this is a useful feature to add. We dont have the hands to implement new features at the moment so this isnt on the horizon but we will take it into account as things move forward.

[RegnerM2015]

Thanks! I may be able to implement via pull request, but I will keep you updated!

[rcorces]

would be very happy to review a pull request!