Barman is an R library for easy data preprocessing and visualization of Genome and Transcriptome single-cell data.
Barman is still under active development, as such installation is done via the actively maintained git repository
devtools::install_github("seanlaidlaw/barman")
Starting from a gene x cell counts matrix barman functions can be used in the following order to achieve a standard workflow:
+------------------------------+
| RNA Counts Matrix |
+--------------+---------------+
|
|
| +-----+
v |
+--------------+---------------+ |
| filter_and_normalise_scrna() | |
+--------------+---------------+ +
| counts
v normalisation
+--------------+---------------+ +
| logR_by_ref_group() | |
+---------+---------+----------+ |
| | |
| | +-----+
| OR |
| | +-----+
v v |
+-------------------------+-+ +--+------------------------+ |
| expression_boxplots() | |get_expression_by_segment()| |
+---------------------------+ +-------------+-------------+ |
| |
| visualisation
v |
+-------------+-------------+ |
| G_T_chr_plot() | |
+---------------------------+ |
|
+-----+
Barman is designed for multiomic data, but provides functions for preprocessing DNA and RNA independently.
Preprocessing the RNA requires a counts matrix, which can be generated from single-cell RNA bams using a tool such as FeatureCounts.
# Load the counts
counts_matrix = read.table("my_featurecounts_matrix.tsv", sep = "\t", header = TRUE)
Filtering can be done automatically using PCA outlier detection, and the resulting filtered matrix is noramlized for gene length using FPKM.
# Filtering and Normalizing
counts_matrix = filter_and_normalise_scrna(counts_matrix) # automatic filtering
Filtering can also be manually specified using the manual_filter option and providing a list of
2 length vectors, specifying the lower and upper limits for total_counts, total_features_by_counts,
%MT genes, and %ERCC respectively.
# Filtering and Normalizing
fpkm_matrix = filter_and_normalise_scrna(counts_matrix, manual_filter = list(c(100000,100000000),
c(1000, 5000), 20, 50)) # filters out cells according to manual filters
We can see the difference in the QC by running the qc_plots function
qc_plots(fpkm_matrix)
We can also normalize based on a given reference group by using the logR_by_ref_group function to
calculate the log fold change between a group of specified reference cells and the rest:
reference_group_cells = rownames(fpkm_matrix)[fpkm_matrix$cell_type = "control",]
logFC_matrix = logR_by_ref_group(fpkm_matrix, reference_group_cells)
We can plot expression per chromosome of different cells by using the expression_boxplots function
which accepts two lists of cell ids, one as an experimental group, and one as a control. From this
it calcualtes the average expression per gene, and groupes by chromosome before plotting.
expression_boxplots(experimental_group = list("Cell_A01", "Cell_A02", "Cell_B03"), control_group =
list("Cell_D03", "Cell_D04", "Cell_D05", fpkm_matrix)
If scCNV has been run on the DNA bams then we can use the produced segmentation files, and a counts matrix to produce segmentation files for the RNA expression data, thus allowing us to compare segment to segment between DNA and RNA.
scCNV_segment_for_A01 = read.table("SEGMENT_scCNV_segment_for_A01.copynumber.refLocus.txt"
A01_expression_segments = get_expression_by_segment("Cell_A01", fpkm_matrix, scCNV_segment_for_A01)
This can also be done in bulk, without having to read in the scCNV produced segment files by using
the bulk_get_expression_by_segment function. this function uses all available cores except one to
process the files in parallel.
bulk_get_expression_by_segment(fpkm_matrix, "./Path_to_scCNV_outputs/", "./RNA_seg_output_folder")
Finally, we can see the Genome and Transcriptome segment plot, showing the DNA to RNA comparison by
running the G_T_chr_plot function.
G_T_chr_plot(cnv_data = scCNV_segment_for_A01, exp_data = A01_expression_segments, title = "Cell A01 Genome and Transcriptome")