Dynamic-deconvolution-algorithm

Dynamic deconvolution algorithm to infer cell type composition and infection-induced states from bulk measurements of complex mixture of immune cell types

System requirements: MATLAB (tested on R2018a)

The code_algorithm.zip contains the following files: README.txt deconvolution_algorithm.m cell_types_marker_genes_from_sc_data.mat small_dataset_to_demo.mat bulk_WT_TLR10_T0_log2_data.mat bulk_WT_TLR10_T4_log2_data.mat bulk_WT_TLR10_T8_log2_data.mat

Instructions: Unzip the dynamic_deconvolution_algorithm.zip file and run the Matlab script from this directory.

Running the code: The function deconvolution_algorithm.m gets as input the name of the .mat file with the bulk RNA-seq data. The function outputs the estimators of each cell type frequency (k) and cell type infection-induced states (s) for each bulk RNA-seq sample. Example to run the code: [k_cell_type_freq,s_cell_type_inf_induced_state]=deconvolution_algorithm('bulk_WT_TLR10_T0_log2_data');

The .mat file with the bulk RNA-seq data should contain 3 variables -

1. exp_data: matrix with the log2 normalized expression data (row for each gene and column for each sample; size: #genes x #samples)
2. exp_genes: cell array with the name of the genes (gene symbols; size: #genes x 1)
3. exp_samples: cell array with the name of the samples (size: #samples x 1)

The code generates two output files:

1. cell_type_freq.txt
2. cell_type_activation.txt

To run the code on a small dataset as demo: Run the deconvolution_algorithm.m script with the input file small_dataset_to_demo.mat, or write as input ‘demo’ instead of the .mat file name. e.g. [k_cell_type_freq,s_cell_type_inf_induced_state]=deconvolution_algorithm('demo') ;

To reproduced the data presented at the manuscript: Run the deconvolution_algorithm.m script with the following input files:

1. bulk_WT_TLR10_T0_log2_data - bulk RNA-seq data of the 8 individuals before infection (t=0)
2. bulk_WT_TLR10_T4_log2_data - bulk RNA-seq data of the 8 individuals 4 hours post-infection (t=4h)
3. bulk_WT_TLR10_T8_log2_data - bulk RNA-seq data of the 8 individuals 8 hours post-infection (t=8h)