sc-marker-genes
This repository holds lists of genes that can be used to identify cell (sub-)types in single cell mRNA-seq experiments.
Objective and goals
The current methodolgy for assigning a putative "cell type" or "subtype" to an individual cell involves (1) embedding cells into a low dimensional manifold using their high dimensional gene expression profiles, (2) clustering cells with similar low dimensional expression profiles together, (3) extracting "marker" genes whose expression is limited to specific clusters or subclusters, and finally (4) associating a known cell (sub-)type to all cells in that cluster based on biological knowlege of those marker genes.
Each of the steps above (and several prior preprocessing steps) rely on thresholds, parameter tuning, and some level of biological insight. An alternative is to train a machine to identify the identity of a cell using a large collection of already-labeled cells. Projects to construct such classification methods are underway or already exist for specific organs, tissue types, or biological conditions, [1][2][3] as examples.
Training a classifier needs high quality cell (sub-)type representatives, and to amass such training data, robust cell (sub-)type candidates must first be identified by hand by the expression of marker genes specific to those cell (sub-)types. The goal of this repository is to open/crowd-source marker genes which can be used to robustly identify cell types.
Contributing and formats
Contribute via pull requests a la Sean Davis's awesome-single-cell.
In an attempt to standardize cell types and gene names, I will enforce the use of Cell Ontology (CL) IDs for celltypes. Find CL IDs for cells at the OLS or Ontobee.
Marker gene list formats are not finalized. For the time being, the format
will consist of a single plain .csv file for each cell type with the below
formatting. This allows easy viewing on GitHub, copy/paste, and concatentation,
as well as most dataframe parsers support by default and it's easy to use with
Excel/OpenOffice.
Filename formating
File format is as follows:
CL_ID.cell_type_name.csv
For example, take B cells (CL_0000236):
CL_0000236.b_cell.csv
This is distinct from say a mature B cell (CL_0000785):
CL_0000785.mature_b_cell.csv
Due to the hierarchical nature of the ontologies, any markers for mature B cells are necessarily B cell markers, so when choosing where to add a marker gene, add it as deep in the onotology as you are comfortable (and can hopefully substantiate).
List formatting
The list itself will take the following headerless format, with one gene per line:
Marker gene,Species,ENSGID,Comment,Reference (DOI/URL)
Taking B cells as an example again, I may add the following:
# file: CL_0000875.b_cell.csv
MS4A1,homo sapiens,ENSG00000156738,doi:10.1038/ncomms14049
A more complicated case might be monocytes:
# file: CL_0001054.monocyte.csv
CD14,homo sapiens,ENSG00000170458,CD14 monocyte (CL_0001054),doi:10.1038/nri3158
FCGR3A,homo sapiens,ENSG00000203747,CD16 monocyte (CL_0002396),doi:10.1016/j.immuni.2010.09.007
Please avoid tabs and quotes.
If a certain subtype has some marker genes above some critical threshold yet to be determined we can then split it off into its own file.
Genes present in multiple species should be added one per line with the correct ENSGID for that species. I don't want to break species up into different files because ultimately I'd like to learn relationships between species.