
Identify cell niches from NMF latent factors
compute_cells_niches.RdFor each NMF factor, selects the cell groups that contribute most to it (the top_n highest weights of the
factor in W, see extract_contributing_features()) and describes them by their deconvolution features
and CCA coefficients. Deconvolution features are kept as they are (subgroups are not expanded into their members).
Usage
compute_cells_niches(
latent_factors,
cell_groups,
top_n = 5,
return = TRUE,
file_name = NULL
)Arguments
- latent_factors
A list returned by
compute.latent_factors()(itsWis used).- cell_groups
The cell groups, as returned by
construct_cell_groups()(itsWeightsare used).- top_n
Integer. Number of cell groups selected per factor. Default 5.
- return
Logical. If
TRUE(default), saves the networks inResults/Cells_niches_<file_name>.pdf(Results/Cells_niches.pdfwithoutfile_name), one page per factor, and a heatmap of the NMF weights (W) of all the cell groups (rows: cell groups, ordered by the factor where their weight is highest and by decreasing weight; columns: factors; outlined: the top cell groups of each factor) inResults/Cells_niches_heatmap_<file_name>.pdf.- file_name
Character. Suffix appended to the output file name.
Value
A data frame with one row per factor, selected cell group and feature: factor, cell_group,
module (TF module of the cell group), weight (NMF weight of the cell group in the factor),
feature and coefficient (CCA coefficient of the feature in the cell group score). Rows are ordered by
factor and by decreasing weight.
Details
With return = TRUE, a tree network of each factor is saved, one page per factor: the factor on the left,
its top cell groups in the middle (filled with the colour of their TF module) and the deconvolution features of each
cell group listed from top to bottom on the right, next to their cell group. The edge between a cell
group and a feature is the CCA coefficient of the feature in the cell group score (xcoef in Weights,
on the z-scored features): red if positive (the score increases with the feature), blue if negative, and wider for
a larger absolute coefficient (coefficients close to zero are hairlines).
The factors describe the high-activity side of the TF modules of their cell groups (the NMF input is the positive part of the cell group scores). A red edge therefore means that the feature is higher in samples where the cell group is active, a blue edge that it is lower. The coefficients of features correlated with each other (e.g. several estimates of the same cell type) are estimated jointly, so a single coefficient can be small or have the opposite sign of the others of its cell group even when the feature itself is related to the TF module like them: read the sign of a single edge together with the rest of its cell group.
Examples
data("network.tuto")
data("deconv_subgroups.tuto")
cell_groups <- construct_cell_groups(network.tuto, deconv_subgroups.tuto, return = FALSE)
#> 44 of 897 features were not assigned to any cluster by the tree cut and are not part of any cell group (pink: 2, yellow: 8, red: 2, blue: 6, brown: 4, turquoise: 12, black: 4, green: 6)
latent <- compute.latent_factors(cell_groups$Cell_groups, return = FALSE)
#> Selecting the rank with bootstrap NMF (100 runs per rank, ranks 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)...
#> Selected rank: 5 (largest rank with all factors robust and none redundant)
#> Running RcppML NMF with rank = 5...
niches <- compute_cells_niches(latent, cell_groups, return = FALSE)
head(niches)
#> factor cell_group module weight
#> 1 Factor1 Dendrogram_red_group_3 red 0.0222516
#> 2 Factor1 Dendrogram_red_group_3 red 0.0222516
#> 3 Factor1 Dendrogram_red_group_3 red 0.0222516
#> 4 Factor1 Dendrogram_red_group_3 red 0.0222516
#> 5 Factor1 Dendrogram_red_group_3 red 0.0222516
#> 6 Factor1 Dendrogram_red_group_3 red 0.0222516
#> feature coefficient
#> 1 B.cells_Subgroup.5 -0.0219763701
#> 2 Neutrophils_Subgroup.3 -0.0091728791
#> 3 DeconRNASeq_LM22_NK.resting -0.0160350224
#> 4 DeconRNASeq_CBSX.NSCLC.PBMCs.scRNAseq_NKT.cells 0.0003870828
#> 5 DeconRNASeq_CBSX.NSCLC.PBMCs.scRNAseq_CD4.cells 0.0014083163
#> 6 DeconRNASeq_BPRNACan_CD8.cells -0.0396184688