scEntropy

We propose scEntropy method and scEGMM framework(see [1]) to measure the order of the cellular transcriptome profile from single cell RNA-seq data.

Getting started

scEntropy method is a reference based method to obtain scEntropy index for each cell in a data set. We provide two different options to generate scEntropy. One is to use predefined reference cell expression(RC), and then calculate scEntropy index for each cell relative to RC. The other is to apply scEGMM framework to automatically identify reference cell expression in a data set and accordingly calculate scEntropy for each cell.

Here are the examples in which we use a data set which contains thousands of cells from 18 head and neck squamous cell carcinoma(HNSCC) patients. The HNSCC data set can be downloaded from https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE103322. The raw data is processed first to obtain clean gene expression data. The details of data pre-processing are shown in sample_code.py.

The examples of two different scEntropy options are given as follows:

scEntropy with predefined reference

# You need to use `pip install git+https://github.com/jzlei/scEntropy.git` to install this package
import scEntropy.scEntropy as scEntropy 
import pandas as pd
import os
data_file_path = os.path.join('data', 'HNSCC_gene_expression.csv')
df_data = pd.read_csv(data_file_path)
scEntropy.scEntropy(df_data, ref_vec=None, option='predefined')

If you don't pass any ref_vec as reference cell(RC), the predefined procedure will use the average gene expression of all cells to replace the ref_vec. But if you pass a ref_vec which is not None type, the predefined procedure will use this passed reference cell to generate scEntropy index.

scEGMM framework

# You need to use `pip install git+https://github.com/jzlei/scEntropy.git` to install this package
import scEntropy.scEntropy as scEntropy
import pandas as pd
import os
data_file_path = os.path.join('data', 'HNSCC_gene_expression.csv')
df_data = pd.read_csv(data_file_path)
scEntropy.scEntropy(df_data, option='RCSA')

If you use scEGMM framework, you don't need to pass any reference cell, since the scEGMM framework will automatically impose the RCSA to identify the intrinsic reference cell in a data set. And then scEntropy index will be generated.

To see the details of the above two different samples, please check the function scEntropy_with_pre_ref and scEntropy_with_RCSA in sample_code.py.

Final words

Here are some useful information to install the scEntropy package and check details of scEntropy method.

Installation:

pip install git+https://github.com/jzlei/scEntropy.git

Requirements:

• Numpy
• Scipy
• Pandas
• Scikit-Learn

Papers:

• [1] Single-cell entropy to quantify the cellular transcriptome from single-cell RNA-seq data