CellCover

This is the python version of CellCover. To run CellCover, Gurobi installation is necessary. Please follow the README.md in lanlanji/CoveringPackage (github.com) to acquire and install the Gurobi academic liscence.

Installing CellCover:

pip install CellCover

Using CellCover to obtain marker panel

To run the CelCover, the following python variables need to be define first:

• data: a numpy array of your single cell RNA-seq data with shape (N,G) where N is the number of cells and G is the size of the gene portfolio.

• gene: a numpy array of gene names with shape (G,)

• CellTypeLabels: a numpy array of cell types with shape (N,)

• CellTypeNames: a numpy array of distinct cell type names (string)

• ct: a single string of the cell type name that user want to find covering markers for, e.g. "CD4"

What is more, there is a list of hyperparameters that need to be defined before running the covering:

• binarization_threshold: the threshold above which we binarize the gene expression to 1, below which we binarize the gene expression to 0

• minSize: the depth of covering

• alpha: 1 - covering rate. The default is $0.05$

• te: This is a parameter for pruning the data. For each cell type, the gene expressing more than te * 100 percent of time are selected for finding the covering markers. The default is $0.1$.

• top_num_gene: This is another pruning parameter. In each class, top_num_gene number of genes with the highest margin score will be selected for marker selection. The default is $6000$.

The pipeline of getting the covering marker panel of the user defined cell type ct is

from CellCover import binarization
from CellCover import SensList
from CellCover import weight
from CellCover import covering
from CellCover import getCoveringVariables
data = binarization(mat = data, binarization_threshold =  binarization_threshold)
sens = SensList(mat = data, CellTypeLabels =  CellTypeLabels, CellTypeNames=CellTypeNames)
X,w,g = weight(mat = data,sens =sens, CellTypeLabels = CellTypeLabels, CellTypeNames = CellTypeNames, ct = ct,gene = gene)
cov = covering(Z=X, minSize = minSize, alpha = alpha,weights = w)
marker = getCoveringVariables(cov, ngenes = len(g), geneNames = g, nlevels = 1)