We want to identify cell-types that are enriched for both phenotype (e.g. cell phenotype) and relate to the external information. Graph-based approaches for identifying these modules can suffer in the single-cell setting because there is an extremely large number of cells profiled per sample and we often profile multiple samples with multiple different experimental conditions or timepoints. Here is Cytocoarsening github link.
If you'd like to install by PYPI, please type this line on your command line(Cytocoarsening PYPI):
pip install cytocoarsening
- Or you can clone the git repository by,
git clone https://github.com/ChenCookie/cytocoarsening.git
- Once you've clone the repository, please change your working directory into this folder.
cd cytocoarsening
If cytocoarsening appears error or warning, please make sure the version of scipy and networkx:
scipy==1.6.2
networkx==2.6.2
To reinstall the particular version of package
pip install --force-reinstall scipy==1.6.2
pip install --force-reinstall networkx==2.6.2
Take preeclampsia for example, if you'd want to list all of the publicly available files for download,
from lxml import html
import requests
r = requests.get(f'https://zenodo.org/record/6779483#.Yrygu-zMJhF')
content = html.fromstring(r.content)
hrefs = content.xpath('//a/@href')
files = [i for i in hrefs if i.endswith('?download=1')]
files = np.unique(files)
print(files)
If you'd like to download any preeclampsia dataset file from zenodo,
curl 'https://zenodo.org/record/6779483/files/Han-FCS_file_list.xlsx?download=1' --output Han-FCS_file_list.xlsx
The function can be excute at one line.
coarsening_group,group_edge,result_dicts=cytocoarsening(cell_data,cell_label,multipass,k_nearest_neighbors)
input
cell_data
- numpy.ndarray. The single cell data with several features. The shape of ndarray is (cell number,features number). Default:cell_data = None
.cell_label
- numpy.ndarray. The attribute of each cell data. The shape of ndarray is (cell number,). Default:cell_label = None
.multipass
- int. The pass number that what want the data size decrease. Default:multipass = 10
.k_nearest_neighbors
- int. Number of neighbors in the inisial graph in each pass. Default:k_nearest_neighbors = 5
.
output
coarsening_group
- dict. The dictionary that indicate supernode as key and the node number list of the group as value in coarsening graphgroup_edge
- numpy.ndarray. The array that record the edge that combine two nodesresult_dicts
- dict. The dictionary that save different result value, including accuracy, error rate, quadratic equation evaluation in feature and label, node number, edge number, runtime, and keypoint
from cytocoarsening.cytocoarsening import cytocoarsening
import numpy as np
import random
cell_data=[[random.random() for i in range(33)] for j in range(4500)]
cell_data=np.array(cell_data)
cell_label = np.array([0] * 1000 + [1] * (3500))
np.random.shuffle(cell_label)
group,edge,diccts=cytocoarsening(cell_data,cell_label,3,5)