Perform geodesic pairing using ATAC/RNA co-embedding components
Usage
cell_pairing(
ATACpcs,
RNApcs,
mode = "geodesic",
tol = 1e-04,
search_range = 0.2,
max_multimatch = 5,
umap_knn_k = 30,
min_subgraph_size = 50,
cca_umap_df = NULL,
seed = 123
)Arguments
- ATACpcs
combined co-embedding components matrix of the ATAC cells. Must have valid ATAC cell barcode names as the rownames.
- RNApcs
combined co-embedding components matrix of the RNA cells. Must have valid RNA cell barcode names as the rownames (unique from ATAC cell barcodes), and have the same number of components (columns) as the `ATACpcs` matrix
- mode
character specifying the pairing mode. Must be one of 'geodesic' (default) or 'greedy'.
- tol
See
fullmatchfor more information on this parameter- search_range
This determines the size of the search knn window for allowed pairing. search_range * total number of cells = size of knn. Default is 0.2. Increasing this can take more time since we have to evaluate over more possible pairs
- max_multimatch
Maximum number of cells in the larger dataset that is allowed to be matched to each cell in the smaller dataset (after up-sampling). Default is 5. This is only to allow for a solvable optmatch solution given geodesic constraints
- umap_knn_k
Number of geodesic ATAC x RNA neighbors to consider in co-embedding graph
- min_subgraph_size
Minimum number of cells (ATAC/RNA each) needed for pairing in a given subgraph. Will skip subgraphs with fewer than these cells.Default is 50 cells. Useful to set this to the smallest cell cluster size you might have in the data
- cca_umap_df
optional umap of all ATAC+RNA cells to visualize UMAP of cells from each chunk being paired, colored by subgraph
- seed
numeric specifying seed to use for subgraph UMAP and for down-sampling in case subgraphs are imbalanced