wot 0.5.4

Optimal transport for time-course single cell data

WOT: Waddington-OT

Waddington-OT uses time-course data to infer how the probability distribution of cells in gene-expression space evolves over time, by using the mathematical approach of Optimal Transport (OT).

Install

Waddington-OT depends on Python 3.

Dependencies

You can install dependencies for wot with conda :

conda install numpy pandas h5py cython scikit-learn scipy matplotlib
conda install -c conda-forge pot

Install the wot package

pip install --user wot

Usage

Initializing an OT Model

wot uses an OTModel as its interface for computing transport maps.

You can initialize an OT Model in python with :

ot_model = wot.ot.initialize_ot_model('matrix.txt', 'days.txt')

All Optimal Transport parameters can be customized when initializing the model. For instance, you could explicitely specify the defaults :

ot_model = wot.ot.initialize_ot_model('matrix.txt', 'days.txt', tmap_prefix='tmaps',
    epsilon=.05, lambda1=10, lambda2=50, batch_size=50, tolerance=1e-2)

You can compute all transport maps with :

ot_model.compute_all_transport_maps()

Loading Transport Maps

Once the transport maps have been created, you can operate on the transport maps using the TransportMapModel interface :

tmap_model = wot.tmap.TransportMapModel.from_directory('.')

All previously computed transport maps will be available.

Changing parameters

If you want to keep the previously computed transport maps, simply initialize a new model with a different prefix. Any model will only affect files that use its tmap_prefix, there is no interaction between models with different prefixes.

Using wot.commands

All data-processing functions are located in the wot.commands subpackage. These include :

• ancestor census
• gene set scores
• gene regulatory networks (grn)
• local enrichment
• optimal transport validation
• trajectory
• trajectory trends
• convert matrix
• force layout
• wot server (interactive version of wot)

All of these are documented on wot's github pages website, with examples using simulated data to show how to use and plot the results of these commands.

Documentation

The full documentation for wot is available on Github Pages : http://broadinstitute.github.io/wot

For more advanced usage, you may also browse the source code to read each function's documentation. Most of wot's internal functions have docstrings with a description of their parameters, output and examples on how to use them.

Developer Notes

For more information about the internal functionning of wot, please refer to the Developer Notes