https://github.com/kieranrcampbell/ouijaflow

Probabilistic single-cell pseudotime with Edward+Tensorflow
https://github.com/kieranrcampbell/ouijaflow

Last synced: about 2 months ago
JSON representation

Probabilistic single-cell pseudotime with Edward+Tensorflow

• Host: GitHub
• URL: https://github.com/kieranrcampbell/ouijaflow
• Owner: kieranrcampbell
• Created: 2017-10-05T14:27:30.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2017-10-05T17:08:56.000Z (about 7 years ago)
• Last Synced: 2024-05-21T23:06:35.422Z (7 months ago)
• Language: Python
• Size: 6.84 KB
• Stars: 12
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# ouijaflow

`ouijaflow` implements the probabilistic single-cell pseudotime model Ouija in Edward and Tensorflow, allowing scalable inference on large single-cell datasets. Inference is performed using reparametrization gradient variational inference.

# Getting started

At present `ouijaflow` may be in stalled via

```

pip install git+https://github.com/kieranrcampbell/ouijaflow.git

```

Fitting pseudotimes with Ouijaflow is straightforward, following the `sklearn` syntax. If `Y` is a cell-by-gene numpy array of non-negative log expression values, then the pseudotimes may be fit via

```python

from ouijaflow import ouija

oui = ouija()

oui.fit(Y)

```

The pseudotimes can be extracted using the `trajectory` function:

```python

z = oui.trajectory()

```

The gene-specific behaviour may be extracted using the `gene_behaviour` function:

```python

oui.gene_behaviour()

```

which returns a pandas data frame with interpretable gene parameters as explained in `oui.gene_behaviour.__doc__`.

For more fine-grained control over the posterior distributions, the approximating distributions may be retrieved in a dictionary using

```python

oui.approx_dists()

```