Test deconvolution methods: CIBERSORTx

[sjspielman]

Broken out of #126

This issue tracks exploring usage of CIBERSORTx for bulk cell type deconvolution. We should not run with an scRNA-seq reference for this exploration. Sources:

• Paper: https://doi.org/10.1038/s41587-019-0114-2
• Methods paper: https://pmc.ncbi.nlm.nih.gov/articles/PMC7695353/
• Website: https://cibersortx.stanford.edu/

[sjspielman]

Noting that these are the cell types in* the 2015 CIBERSORT paper's signature gene matrix, as copied from the paper's table S1 (attached), though it is possible to create one's own signature matrix. Their matrix has 22 flavors of leukocyte, including both cell types and states, but no other "normal" cell types like fibroblasts, epithelial, etc.
41592_2015_BFnmeth3337_MOESM207_ESM.xls

B cells naive	B cells memory	Plasma cells	T cells CD8	T cells CD4 naive	T cells CD4 memory resting	T cells CD4 memory activated	T cells follicular helper	T cells regulatory (Tregs)	T cells gamma delta	NK cells resting	NK cells activated	Monocytes	Macrophages M0	Macrophages M1	Macrophages M2	Dendritic cells resting	Dendritic cells activated	Mast cells resting	Mast cells activated	Eosinophils	Neutrophils

[allyhawkins]

Noting that these are the cell types in* the 2015 CIBERSORT paper's signature gene matrix, as copied from the paper's table S1 (attached), though it is possible to create one's own signature matrix. Their matrix has 22 flavors of leukocyte, including both cell types and states, but no other "normal" cell types like fibroblasts, epithelial, etc. 41592_2015_BFnmeth3337_MOESM207_ESM.xls

B cells naive B cells memory Plasma cells T cells CD8 T cells CD4 naive T cells CD4 memory resting T cells CD4 memory activated T cells follicular helper T cells regulatory (Tregs) T cells gamma delta NK cells resting NK cells activated Monocytes Macrophages M0 Macrophages M1 Macrophages M2 Dendritic cells resting Dendritic cells activated Mast cells resting Mast cells activated Eosinophils Neutrophils

Just a thought that is not fully baked, but if you can create a "custom signature", can we use the celldex references that we use for SingleR to create a signature that would be used for deconvolution here? Then you would be looking at the same cell types that we are able to identify in single-cell. Again, not fully baked because I haven't read about this as thoroughly as you, but maybe it's something we could think about?

[sjspielman]

Again, not fully baked because I haven't read about this as thoroughly as you, but maybe it's something we could think about?

Unfortunately, I too have not read this thoroughly because the PDF version of the paper I have doesn't include the paywalled "Online Methods" section which describes how the signature matrix was created in the first place. If we want to to pursue this, step 1 is asking a friend for the full paper 😬

[jaclyn-taroni]

Why is this insufficient? https://pmc.ncbi.nlm.nih.gov/articles/PMC4739640/#S1

[sjspielman]

Why is this insufficient?

It is highly sufficient! I just forgot it existed...thanks!