UXM fragment-level deconvolution algorithm

January 4, 2023 · View on GitHub

UXM is a computational fragment-level reference-based deconvolution algorithm for DNA methylation sequencing data. It constructs a reference atlas where the percentages of unmethylated fragments is computed for every marker (row) in each cell type (column). A non-negative least squares (NNLS) algorithm is then used to fit an input sample, and estimate its relative contributions.

Visualization of the reference atlas published in Loyfer et al.

Quick start

Installation

Make sure wgbstools is installed and found in $PATH.
Clone UXM.
Optionally, add the "uxm" main scrip to your $PATH.

git clone https://github.com/nloyfer/UXM_deconv.git
cd UXM_deconv/
# Optionally, add to PATH (or link "./uxm" to some directory in $PATH)
export PATH=${PATH}:$PWD

Dependencies

python 3+
pandas, numpy, scipy
wgbstools

Usage examples

Running deconvolution

$ uxm deconv data/*pat.gz -o output.csv
dumped atlas to output.csv

Plotting deconvolution results

# Plot a stacked bar plot
$ uxm plot output.csv -o output.pdf
dumped figure to output.pdf

This project is developed by Netanel Loyfer in Prof. Tommy Kaplan's lab at the Hebrew University, Jerusalem, Israel.
If you are using the UXM deconvolution in a paper, please cite:

Loyfer*, Magenheim*, Peretz* et al. (2023) ‘A DNA methylation atlas of normal human cell types’, Nature.