alexmascension / triku_notebooks

Home Page:https://gitlab.com/alexmascension/triku

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

triku notebooks

In this repo we are going to upload the notebooks related to the triku package (https://gitlab.com/alexmascension/triku). The notebook will contain information about the package and the underlying rationale, as well as the code necessary to reproduce the figures in the manuscript.

If the figures after production do not look exactly like in the paper, that is because cosmetic changes were done. None of the figures have been subjected to data manipulation.

The notebooks are mean to be run in the order they appear (first 0, then 1, etc.). Notebooks starting by X do not provide information of figures appearing in the paper, but are nonetheless interesting to run. The contents of the notebooks are the following:

  • 0A_Environment_and_data_download: This is the first notebook. After creating the environment (see How to run the notebooks), it installs complementary R packages and downloads / processes the required data for the analysis.
  • 0B_Generation_of_artificial_datasets: It creates random artificial datasets using splatter. It produces the Figure S5
  • 1_Comparing_feature_selection_methods: Reproduction of the main paper analysis. It produces the Figures 3, 4, 5, 6, 7 and Figures S1, S2, S3, and S4.
  • 2A_Reads_in_knn_cells_as_a_method_for_feature_selection: This is the fundational notebook with the rationale of triku. It includes Figures S6 and S7
  • 2B_Triku_stability_measures: This notebook shows the robustness of the parameters of triku. It includes the Figure S8.
  • Notebooks starting with X are notebooks in development, or with results not relevant for the paper.

How to run the notebooks

If you want to reproduce the analysis in the notebooks, the best way to do it is this. We recommend that you use conda (miniconda or anaconda) as your package manager. We also recommend to install mamba, a package that works as conda, but optimizes the package download and resolution of dependencies. Thus,

  • First download miniconda3 from here.
  • Then, in the base environment, install mamba with conda install -c conda-forge mamba
  • To ensure reproducibility, create a new environment with the yaml file: mamba env create -f condaenv_triku_notebooks.yaml

If you want to see the output of the notebooks to replicate the figures, you can download them from our Zenodo repository.

Notes

This project comes from the triku repository. We have exported the notebooks to this second repository to keep the development of the software and the notebooks separate. The commit history of the notebooks before the separation can be consulted there.

Some UMAPs or lesser parts of the figures might differ. Although seeds are set, some parts of the analysis seem to be variable to minute changes in the data. Although this happens, the interpretation of the results is still the same.

About

https://gitlab.com/alexmascension/triku

License:BSD 3-Clause "New" or "Revised" License


Languages

Language:Jupyter Notebook 79.8%Language:Python 20.2%