CIBERSORTx requires the use of marker gene reference profiles to enumerate cell subsets in bulk tissue samples. Commonly referred to as a “signature matrix,” marker gene profiles can be derived from a variety of sources, including scRNA-seq data and sorted cell populations profiled by bulk RNA-seq or microarrays [21 (link), 23 (link)]. We previously described a microarray-derived signature matrix for profiling 22 functionally defined human immune cell types, called LM22 [16 (link)]. With scRNA-seq, it is now possible to create a signature matrix that captures all major cell subsets in a tissue without complex sorting experiments (seeNote 1 ). This can enable large-scale investigation of novel or poorly understood phenotypic states in bulk tissue GEPs. In malignancies, cellular states of interest may include subpopulations of activated, resting, or exhausted T cells [24 (link)–26 (link)], cancer-associated fibroblasts [27 (link)], or cancer cells [28 (link), 29 (link)], including tumor initiating cells or cancer stem cells [30 (link)]. As further described below, scRNA-seq can also provide a powerful means of validating reference signatures through the use of mixture samples created from single-cell transcriptomes.
In the next section, we describe how to create a signature matrix from scRNA-seq data. Other platforms are described in detail elsewhere (see tutorials 6 and 7 athttp://cibersortx.stanford.edu/tutorial.php ).
In the next section, we describe how to create a signature matrix from scRNA-seq data. Other platforms are described in detail elsewhere (see tutorials 6 and 7 at