CDSeq: A complete deconvolution method for dissecting tissue heterogeneity
CDSeq is able to estimate both cell-type-specific gene expression profiles and sample-specific cell-type proportions simultaneously using bulk RNA-seq data.
CDSeq MATLAB code version 0.1.1
Free for academic use and educational use. Please contact me at firstname.lastname@example.org for commercial use.
This is the MATLAB code for computational deconvolution method CDSeq.
Reference: A novel computational complete deconvolution method using RNA-seq data (submitted) Authors: Kai Kang, Qian Meng, Igor Shats, David Umbach, Melissa Li, Yuanyuan Li, Xiaoling Li, Leping Li.
Affiliation: National Institute of Environmental Health Sciences, Biostatistics and Computational Biology Branch and Signal Transduciton Lab.
CDSeq MATLAB code version 0.1.2
Added data-dilution option to speed up the algorithm. We provided explaination in the manuscript which you can find on bioRxiv. Check out the demo.m in CDSeq_012 for details.
CDSeq MATLAB code version 0.1.4
- I used unsigned short integer types for large vectors to lower memory cost.
- I added binary search for cell type assignment procedure in Gibbs Sampler. In early version, it was done using linear search.
note: This version of CDSeq takes advantages of new MATLAB features (R2018a and later versions) which allows mex C++ function handles unsigned short int type (2 bytes).
I only compiled the code on Mac and Linux but not Windows yet. Will find a windows machine to compile later. You can also compile by yourself following the instructions below.
To compile CDSeq's C++ function, run the following two lines of commands in MATLAB:
mex -setup C++
mex -R2018a CDSeqGibbsSampler.cpp
Experimental data GEO code: GSE123604
The synthetic data (6 pure cell line GEPs and 40 mixtures) are available upon request since it is too large to put them on github.