jj_association_heatmap: find associations of embedding scores with meta data
In mathosi/jj: Just Joise

View source: R/jj_plot_association_heatmap.R

jj_association_heatmap

R Documentation

find associations of embedding scores with meta data

Description

jj_metadata_association_heatmap: Plot heatmap to find metadata associations with a set of scores, e.g. from PCA jj_association_heatmap: Plot mean or scaled mean per group for all columns in scores_mat, good to further understand single metadata associations with scores. Further arguments can be passed to ComplexHeatmap jj_test_associations: Function to test association of metadata with a group of interest. Run logistic regressions with group1/group2 as outcome and a column in the metadata_df as predictor Repeat for all columns and sort by p-value and deviance against a null model (group ~ 1). This can reveal metadata variables enriched for the cluster of interest (group1)

Usage

jj_metadata_association_heatmap(
  score_mat,
  metadata_df,
  plotAssocFor = NULL,
  cor_cutoff = 0.2,
  categorical_skip = 30,
  ...
)

Arguments

`metadata_df`	data.frame/matrix with the metadata for each cell
`plotAssocFor`	Option to specify subset of metadata_mat for which associations should be calulated
`cor_cutoff`	Only show correlations with absolute value above this cutoff for better overview
`categorical_skip`	Skip categorical metadata columns with more than this number of distinct values
`scores_mat`	numeric matrix of scores (eg pc scores)
`group_vec`	vector of length = nrow(scores_mat) with group annotation
`group1`	cell names of group, for which jj_test_associations should test associations. Used to select the subset of metadata_df (which needs to contain rownames with the cell names
`group2`	group1 is compared against group2. If group2=NULL (default), all cells in the metadata_df not part of group1 are defined as group2

Examples

#e.g. PC2 has positive correlation with the grouping from RNA_snn_res.1
pca_scores = Embeddings(pbmc_small, 'pca')[,1:10]
jj_metadata_association_heatmap(score_mat = pca_scores, metadata_df = pbmc_small[[]],
                                cor_cutoff = 0.2)
# Groups 1 and 2 tend to have high scores and group 0 has low scores for PC2
jj_association_heatmap(scores_mat = pca_scores, group_vec = pbmc_small$RNA_snn_res.1)
# confirm by plotting the data
jj_plot_features(pbmc_small, reduction = 'pca', meta_features = 'RNA_snn_res.1', pt.size = 2)
#test association of meta data with one selected group of cells
group1 = colnames(pbmc_small)[pbmc_small$RNA_snn_res.1 == '0']
jj_test_associations(metadata_df = pbmc_small@meta.data, group1 = group1)

mathosi/jj documentation built on Feb. 25, 2024, 2:29 p.m.