SnapATAC (version 1.0.0)

runDiffusionMapsExtension: Diffusion Maps Extension

Description

This function takes two snap objects - one for reference dataset and one for query dataset and computes the diffusion maps embedding for the query dataset by projecting the query cells into the pre-computed diffusion.

Usage

runDiffusionMapsExtension(obj1, obj2, input.mat = c("bmat", "pmat"))

Arguments

obj1

A snap obj for reference dataset

obj2

A snap obj for query dataset

input.mat

Input matrix c("bmat", "pmat").

Details

The computational complexity of diffusion maps algorithm exhibits quadratic growth with the increase of cells, making it infeasible for large-scale datasets. To overcome this limitation, we apply Nystrom landmark diffusion map algorithm to efficiently generate the low-dimension embedding for large-scale dataset. A practical Nystrom landmark diffusion map algorithm project the query dataset onto the low-dimensional embedding space as learned from the refernce dataset to create a embedding space for query cells.

Examples

# NOT RUN {
data(demo.sp);
demo.sp = makeBinary(demo.sp);
# }