semir2 / cibersortx Goto Github PK

IF each column of 'single cell reference matrix' is a phenotype label, proceed to step 2

<From cibersortx website>  
The reference sample file is an input file required for custom signature genes file generation by CIBERSORT and consists of a table of the gene expression profiles of reference sample cell populations that will be compared to each other as defined in the phenotype classes file (next section) to generate the custom signature genes file. Each column corresponds to the gene expression profile of a reference sample. Multiple replicates of a given reference sample cell type (one in each column) may be used.'

ELSE take the following two files and convert them using 'convert_reference_file.py' single cell reference matrix class matrix

python convert_reference_file.py MCA_liver_cell_expression.tsv MCA_liver_cell_class.tsv

⭐ parameters (pro tem)
Single cell input options:
Min. Expression 0.5
❗ Download output signature matrix

Filter out columns with no expressions, using the signature matrix from step 2

python filter_by_signature_genes.py signature_gene_file.txt mixture_file.tsv

This creates a new mixture file named {mixture_file}_filtered.tsv

Upload the Mixture matrix from step 3 to the Cibersortx website, then run fraction imputation
⭐ parameters
default

💫 Memo
For the mouse liver data set, 19133+ samples exceeded the 'Allowed memory size of 536870912 bytes'
Use something like the following command to divide data set

# first 10000 samples
cat mixture_file_filtered.tsv  |  cut -f 1-10001  > mixture_file_filtered_1.tsv
# the rest of the samples (if <10000); first column == gene names (index)
cat mixture_file_filtered.tsv  |  cut -f 1,10002-  > mixture_file_filtered_2.tsv

And to check the number of columns per file

cat file.tsv | awk '{print NF}'| sort -nu | tail -n 1  

For validating Cibersortx's results
Using single cell data, mix two cell types together and run cibersortx.

The below command generates an tsv file which may be used as cibersortx input.
Each column consists of a 50%-50% mix of two cell type expressions

python validation.py single_cell_reference_made_in_step_1.tsv

Use the following Jupyter Notebook. Input file will be a Cibersortx job result file (Default in .txt format).

Sort_Cibersortx_output_fractions.ipynb