This repository contains the corresponding code for the Master's Thesis "Overcoming Sparsity and Technical Noise: Systematic Evaluation of Imputation and Denoising Methods for Single-Cell Metabolomics and Lipidomics Data" by Marius Klein and additional functions for the future analysis of single-cell metabolomics and lipidomics data.
The Python package can be downloaded from this GitHub repo via pip. Just run this command in your Terminal:
$ pip install git+https://github.com/mariusrklein/sc_imputation_denoising
This installs the package and all dependencies.
The package contains the following modules:
- imputation: contains functions for the preparation of the data for imputation or denoising, e.g. filtering of sparse ions and cells, simulation of missing values using MCAR or MNAR mechanisms, and the calculation of the missing value rate.
- evaluation: contains functions for the high-level evaluation of imputation and denoising methods and the visualization of the respective results, e.g. collective calculation of the evaluation metrics, the complete workflow of sparsity simulation, imputation, and evaluation, visualization of the results
- metrics: implementation of the individual metrics employed in the work, e.g. mean squared error of values (MSE), cluster separation metrics.
Examples of the usage of the functions are provided in this tutorial.