Single-Cell Interpretable Tensor Decomposition

Documentation for package ‘scITD’ version 1.0.4

DESCRIPTION file.

Help Pages

-- A --

Apply ComBat batch correction to pseudobulk matrices. Generally, this should be done through calling the form_tensor() wrapper function.

-- C --

calculate_fiber_fstats	Calculate F-Statistics for the association between donor scores for each factor donor values of shuffled gene_ctype fibers
check_rec_pres	Helper function to check whether receptor is present in target cell type
clean_data	Clean data to remove genes only expressed in a few cells and donors with very few cells. Generally, this should be done through calling the form_tensor() wrapper function.
colMeanVars	Calculates column mean and variance. Adapted from pagoda2. https://github.com/kharchenkolab/pagoda2/blob/main/src/misc2.cpp
compare_decompositions	Plot a pairwise comparison of factors from two separate decompositions
compute_associations	Compute associations between donor proportions and factor scores
compute_donor_props	Get donor proportions of each cell type or subtype
compute_LR_interact	Compute and plot the LR interactions for one factor
convert_gn	Convert gene identifiers to gene symbols
count_word	count_word. From older version of simplifyEnrichment package.

-- D --

determine_ranks_tucker

Run rank determination by svd on the tensor unfolded along each mode

-- F --

form_tensor

Form the pseudobulk tensor as preparation for running the tensor decomposition.

-- G --

get_all_lds_factor_plots	Generate loadings heatmaps for all factors
get_callouts_annot	Get gene callout annotations for a loadings heatmap
get_ctype_exp_var	Get explained variance of the reconstructed data using one cell type from one factor
get_ctype_prop_associations	Compute and plot associations between donor factor scores and donor proportions of major cell types
get_ctype_subc_prop_associations	Compute and plot associations between donor factor scores and donor proportions of cell subtypes
get_ctype_vargenes	Partition main gene by cell matrix into per cell type matrices with significantly variable genes only. Generally, this should be done through calling the form_tensor() wrapper function.
get_donor_meta	Get metadata matrix of dimensions donors by variables (not per cell)
get_factor_exp_var	Get the explained variance of the reconstructed data using one factor
get_fstats_pvals	Calculate adjusted p-values for gene_celltype fiber-donor score associations
get_gene_modules	Compute WGCNA gene modules for each cell type
get_gene_set_vectors	Get logical vectors indicating which genes are in which pathways
get_indv_subtype_associations	Compute subtype proportion-factor association p-values for all subclusters of a given major cell type
get_intersecting_pathways	Extract the intersection of gene sets which are enriched in two or more cell types for a factor
get_leading_edge_genes	Get the leading edge genes from GSEA results
get_lm_pvals	Compute gene-factor associations using univariate linear models
get_max_correlations	Computes the max correlation between each factor of the decomposition done using the whole dataset to each factor computed using the subsampled/bootstrapped dataset
get_meta_associations	Get metadata associations with factor donor scores
get_min_sig_genes	Evaluate the minimum number for significant genes in any factor for a given number of factors extracted by the decomposition
get_module_enr	Identify gene sets that are enriched within specified gene co-regulatory modules. Uses a hypergeometric test for over-representation. Used in plot_multi_module_enr().
get_normalized_variance	Get normalized variance for each gene, taking into account mean-variance trend
get_num_batch_ranks	Plot factor-batch associations for increasing number of donor factors
get_one_factor	Get the donor scores and loadings matrix for a single-factor
get_one_factor_gene_pvals	Get significant genes for a factor
get_pseudobulk	Collapse data from cell-level to donor-level via summing counts. Generally, this should be done through calling the form_tensor() wrapper function.
get_real_fstats	Get F-Statistics for the real (non-shuffled) gene_ctype fibers
get_reconstruct_errors_svd	Calculate reconstruction errors using svd approach
get_significance_vectors	Get vectors indicating which genes are significant in which cell types for a factor of interest
get_subclusters	Perform leiden subclustering to get cell subtypes
get_subclust_de_hmaps	Get list of cell subtype differential expression heatmaps
get_subclust_enr_dotplot	Get scatter plot for association of a cell subtype proportion with scores for a factor
get_subclust_enr_fig	Get a figure showing cell subtype proportion associations with each factor. Combines this plot with subtype UMAPs and differential expression heatmaps. Note that this function runs better if the number of cores in the conos object in container$embedding has n.cores set to a relatively small value < 10.
get_subclust_enr_hmap	Get heatmap of subtype proportion associations for each celltype/subtype and each factor
get_subclust_umap	Get a figure to display subclusterings at multiple resolutions
get_subtype_prop_associations	Compute and plot associations between factor scores and cell subtype composition for various clustering resolution parameters
get_sums	Calculates factor-stratified sums for each column. Adapted from pagoda2. https://github.com/kharchenkolab/pagoda2/blob/main/src/misc2.cpp

-- H --

ht_clusters

Visualize the similarity matrix and the clustering. Adapted from simplifyEnrichment package. https://github.com/jokergoo/simplifyEnrichment/blob/master/R/ht_clusters.R

-- I --

identify_sex_metadata	Extract metadata for sex information if not provided already
initialize_params	Initialize parameters to be used throughout scITD in various functions
instantiate_scMinimal	Create an scMinimal object. Generally, this should be done through calling the make_new_container() wrapper function.
is_GO_id	Check if a character is a go ID

-- M --

make_new_container	Create a container to store all data and results for the project. You must provide a params list as generated by initialize_params(). You also need to provide either a Seurat object or both a count_data matrix and a meta_data matrix.
merge_small_clusts	Merge small subclusters into larger ones

-- N --

nmf_unfolded	Computes non-negative matrix factorization on the tensor unfolded along the donor dimension
normalize_counts	Helper function to normalize and log-transform count data
normalize_pseudobulk	Normalize the pseudobulked counts matrices. Generally, this should be done through calling the form_tensor() wrapper function.
norm_var_helper	Calculates the normalized variance for each gene. This is adapted from pagoda2. https://github.com/kharchenkolab/pagoda2/blob/main/R/Pagoda2.R Generally, this should be done through calling the form_tensor() wrapper function.

-- P --

parse_data_by_ctypes	Parse main counts matrix into per-celltype-matrices. Generally, this should be done through calling the form_tensor() wrapper function.
pca_unfolded	Computes singular-value decomposition on the tensor unfolded along the donor dimension
plotDEheatmap_conos	Plot a heatmap of differential genes. Code is adapted from Conos package. https://github.com/kharchenkolab/conos/blob/master/R/plot.R
plot_donor_matrix	Plot matrix of donor scores extracted from Tucker decomposition
plot_donor_props	Plot donor celltype/subtype proportions against each factor
plot_donor_sig_genes	Generate a gene by donor heatmap showing scaled expression of top loading genes for a given factor
plot_dscore_enr	Compute enrichment of donor metadata categorical variables at high/low factor scores
plot_gsea_hmap	Plot enriched gene sets from all cell types in a heatmap
plot_gsea_hmap_w_similarity	Plot already computed enriched gene sets to show semantic similarity between sets
plot_gsea_sub	Look at enriched gene sets from a cluster of semantically similar gene sets. Uses the results from previous run of plot_gsea_hmap_w_similarity()
plot_loadings_annot	Plot the gene by celltype loadings for a factor
plot_mod_and_lig	Plot trio of associations between ligand expression, module eigengenes, and factor scores
plot_multi_module_enr	Generate gene set x ct_module heatmap showing co-expression module gene set enrichment results
plot_rec_errors_bar_svd	Plot reconstruction errors as bar plot for svd method
plot_rec_errors_line_svd	Plot reconstruction errors as line plot for svd method
plot_scores_by_meta	Plot dotplots for each factor to compare donor scores between metadata groups
plot_select_sets	Plot enrichment results for hand picked gene sets
plot_stability_results	Generate a plot for either the donor scores or loadings stability test
plot_subclust_associations	Plot association significances for varying clustering resolutions
prep_LR_interact	Prepare data for LR analysis and get soft thresholds to use for gene modules
project_new_data	Project multicellular patterns to get scores on new data

-- R --

reduce_dimensions	Gets a conos object of the data, aligning datasets across a specified variable such as batch or donors. This can be run independently or through get_subtype_prop_associations().
reduce_to_vargenes	Reduce each cell type's expression matrix to just the significantly variable genes. Generally, this should be done through calling the form_tensor() wrapper function.
render_multi_plots	Create a figure of all loadings plots arranged
reshape_loadings	Reshape loadings for a factor from linearized to matrix form
run_fgsea	Run fgsea for one cell type of one factor
run_gsea_one_factor	Run gsea separately for all cell types of one specified factor and plot results
run_hypergeometric_gsea	Compute enriched gene sets among significant genes in a cell type for a factor using hypergeometric test
run_jackstraw	Run jackstraw to get genes that are significantly associated with donor scores for factors extracted by Tucker decomposition
run_stability_analysis	Test stability of a decomposition by subsampling or bootstrapping donors. Note that running this function will replace the decomposition in the project container with one resulting from the tucker parameters entered here.
run_tucker_ica	Run the Tucker decomposition and rotate the factors

-- S --

sample_fibers	Get a list of tensor fibers to shuffle
scale_fontsize	Scale font size. From simplifyEnrichment package. https://github.com/jokergoo/simplifyEnrichment/blob/master/R/ht_clusters.R
scale_variance	Scale variance across donors for each gene within each cell type. Generally, this should be done through calling the form_tensor() wrapper function.
seurat_to_scMinimal	Convert Seurat object to scMinimal object. Generally, this should be done through calling the make_new_container() wrapper function.
shuffle_fibers	Shuffle elements within the selected fibers
stack_tensor	Create the tensor object by stacking each pseudobulk cell type matrix. Generally, this should be done through calling the form_tensor() wrapper function.
stop_wrap	Helper function from simplifyEnrichment package. https://github.com/jokergoo/simplifyEnrichment/blob/master/R/utils.R
subset_scMinimal	Subset an scMinimal object by specified genes, donors, cells, or cell types

-- T --

test_container	Data container for testing tensor formation steps
tucker_ica_helper	Helper function for running the decomposition. Use the run_tucker_ica() wrapper function instead.

-- U --

update_params

Update any of the experiment-wide parameters

-- V --

vargenes_anova

Compute significantly variable genes via anova. Generally, this should be done through calling the form_tensor() wrapper function.