OscoNet

Bootstrap-based OscoNet method: Method to infer sinusoidal oscillations in single cell data.

This software reproduces the approach presented in 'OscoNet: inferring oscillatory gene networks' by Luisa Cutillo, Alexis Boukouvalas, Elli Marinopoulou, Nancy Papalopulu & Magnus Rattray

BMC Bioinformatics volume 21, Article number: 351 (2020)

https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-020-03561-y

contact: [email protected]

The authors would like to thank Michael Croucher @mikecroucher for code optimization.

Installation

1. Create new environment conda create --name fullosconet python=3 and to activate it with conda activate fullosconet before proceeding.
2. Install required packages using pip install -r requirements.txt
3. Install package pip install -e .
4. Install numba conda install numba
5. Verify your installation by running pytest from the project root directory FullOscoNet. Note this can take around 1-2 minutes.

Synthetic data

1. Run python OscopeBootstrap/oscope_tf.py --test_mode for a simple demonstration of the method on synthetic data. This will run under a quick configuration to demonstrate the capabilities of the method. This should take 10-20 seconds.
2. Remove the --test_mode flag for a 1000-sample bootstrap test on the exact synthetic run configuraiton used in the paper (1000 genes, 100 cells with 3 clusters of co-oscillating genes).

Notebooks

1. notebooks/OscoNet introduction.ipynb: provides an introduction to the hypothesis test on a simple synthetic example.
2. notebooks/Reproduce_figures_5_7.ipynb : pseudotime on Whitfield microarray data. To see how the spectral embedding pseudotime method can be applied.
3. notebooks/Reproduce_Table1.ipynb: Reproduce table 1 from OscoNet paper
4. notebooks/Reproduce_Table5.ipynb: Reproduce table 5 from OscoNet paper
5. notebooks/Reproduce_Figure4.ipynb: Reproduce Figure 4 from OscoNet paper

Notes on using real data

OscoNet requires the data to be already normalised and rescaled between [-1,1].

We suggest you chose your favourite normalization pipeline and also you continue the preprocessing in R, using the functions 'MVfilter' and 'NormForSine' from the R package Oscope: https://www.bioconductor.org/packages/release/bioc/html/Oscope.html