Description of the Method

The method CIBERSORT was published in Nature Methods. The original method is described here and the updated method CIBERSORTx was published in Nature Biotechnology and is available here. We have used the original implementaion from immunedeconv, which is a systematic benchmarking study of widely used supervised cell type deconvolution algorithms. Please note, a request needs to be placed to the authors of CIBERSORT before using the Rscript of its function in immunedeconv. We have made the method available in the Docker image¹.

CIBERSORT is a supervised method which uses a curated gene expression signature matrix \((S)\) of 22 or 6 major immune cell types (known as LM22 or LM6 signatures) to estimate the relative abundances \((P)\) of these cell types in any given bulk transcriptomic profiles \((E)\) using linear support vector regression. Essentially, the following eqaution is solved where \(g\) is genes, \(s\) is samples and \(c\) is cell types

\[E_{gs} = S_{gc}.P_{cs}\]