CELLEX (CELL-type EXpression-specificity) is a tool for computing cell-type Expression Specificity (ES) profiles. It employs a "wisdom of the crowd"-approach by integrating multiple ES metrics, thus combining complementary cell-type ES profiles, to capture multiple aspects of ES and obtain improved robustness.

• Documentation
• Quick Start
• Tutorials
• Contact and References

The documentation for CELLEX can be accessed in the following ways:

• CELLEX Wiki : main documentation on the usage of CELLEX
• CELLEX API docs: documentation of CELLEX API/functions
• Pre-print article: technical details on the CELLEX method. Pre-print link: Mapping heritability of obesity by cell types (Timshel bioRxiv, 2020, Supplementary Material)

We are continually updating the documentation for CELLEX. If some information is missing, please submit your request or question via our issue tracker.

This brief tutorial showcases the core features of CELLEX.

import numpy as np
import pandas as pd
import cellex

data = pd.read_csv("./data.csv", index_col=0)
metadata = pd.read_csv("./metadata.csv", index_col=0)

eso = cellex.ESObject(data=data, annotation=metadata, verbose=True)
eso.compute(verbose=True)
eso.results["esmu"].to_csv("mydataset.esmu.csv.gz")

pip install cellex

Clone the development repo and install from source using pip. The development version may contain bug fixes that have not been released, as well as experimental features.

git clone https://github.com/perslab/CELLEX.git --branch develop --single-branch
cd CELLEX
pip install -e .

import numpy as np # needed for formatting data for this tutorial
import pandas as pd # needed for formatting data for this tutorial
import cellex

data = pd.read_csv("./data.csv", index_col=0)
metadata = pd.read_csv("./metadata.csv", index_col=0)

Data may consist of UMI counts (integer) for each gene and cell.

Shape: m genes by n cells.

Metadata should consist of unique cell id's and matching annotation (string).

Shape: n cells by 2.

eso = cellex.ESObject(data=data, annotation=metadata, verbose=True)

eso.compute(verbose=True)

All results are accessible via the results attribute of the ESObject.

eso.results["esmu"]

The ESmu scores may be used with CELLECT. CELLECT requires that genes are in the Human Ensembl Gene ID format. CELLEX provides a simple renaming utility for this purpose:

cellex.utils.mapping.ens_mouse_to_ens_human(eso.results["esmu"], drop_unmapped=True, verbose=True)

eso.results["esmu"].to_csv("mydataset.esmu.csv.gz")

eso.save_as_csv(keys=["all"], verbose=True)

Output consist of Expression Specificity Weights (float) for each gene and cell-type. ESmu values lie in the range [0,1].

Shape: m genes by x unique annotations. N.B. a number of genes may be removed during preprocessing.

Various tutorials and walkthroughs will be made available here, while the Wiki is in the making. These Jupyter Notebooks cover everything from downloading CELLEX and data to analysis and plotting.

• Demo: Downloading and running CELLEX with sample Mousebrain Atlas data
• Demo: Downloading and running CELLEX with sample MOCA data

• Tobias Overlund Stannius (University of Copenhagen) @TobiasStannius
• Pascal Nordgren Timshel (University of Copenhagen) @ptimshel

Please create an issue in this repo, if you encounter any problems using CELLEX. Alternatively, you may write an email to timshel(at)sund.ku.dk

If you find CELLEX useful for your research, please consider citing: Timshel (bioRxiv, 2020): Mapping heritability of obesity by cell types