Aggregating single cell coexpression

Basic functionality

devtools::install_github("PavlidisLab/aggtools")
library(aggtools)

pc_df <- read.delim("/home/amorin/Data/Metadata/refseq_select_mm10.tsv")
dat <- readRDS("/space/scratch/amorin/TR_singlecell/GSE145172/GSE145172_clean_mat_and_meta_CPM.RDS")


result <- aggr_coexpr_single_dataset(mat = dat$Mat, 
                                     meta = dat$Meta,
                                     pc_df = pc_df,
                                     cor_method = "pearson",
                                     agg_method = "FZ",
                                     verbose = TRUE)

Current implementation has strict assumptions about the input data:

1. pc_df is a data.frame with a "Symbol" column whose elements are unique and contain all gene/rownames of mat. Only the elements in pc_df$Symbol are kept: mat will be subset to these genes. An error will be thrown if pc_df$Symbol has genes that are not found in the rownames of mat.

2. meta is a data.frame that has "ID" and "Cell_type" columns -- the latter is the grouping factor for aggregation to subset for IDs/cells in mat. Currently if you want to subset by patient/condition/some other factor other than cell type, you would have to name the corresponding column in meta to "Cell_type".

3. mat is a gene (rows) by cell (columns) sparse matrix of class "dgCMatrix" (see package Matrix)

4. All column names (cell IDs) of mat are found in the "ID" column of meta.

The multi dataset function makes these further assumptions:

5. input_df is a data.frame of single cell datasets containing columns "ID", "Cell_type", and "Path".

6. The data paths specified in input_df$Path lead to .RDS objects of lists with two named elements: "Meta" for the metadata and "Mat" for the sparse count matrix.