vaeda (variaitonal auto-encoder (vae) for doublet annotation (da)) is a python package for doublet annotation in single cell RNA-sequencing. For method details and comparisons to alternative doublet annotation tools, see our pre-print.
You can install vaeda using conda and pip as follows:
conda create -n vaeda_env python=3.8
conda activate vaeda_env
pip3 install --upgrade tensorflow==2.8.0
pip3 install --upgrade tensorflow-probability==0.16.0
pip3 install 'scanpy[leiden]'==1.8.0
pip3 install typing-extensions==3.7.4 absl-py==0.10 six==1.15.0 wrapt==1.12.1 xlrd==1.2.0
pip3 install -i https://test.pypi.org/simple/ vaeda==0.0.30
import vaeda
...
res = vaeda.vaeda(adata)
Where:
- adata is an annotated data matrix with raw counts in adata.X
- res is adata updated with the encoding generated by vaeda stored in adata.obsm['vaeda_embedding'] and the doublet scores and calls stored in adata.obs['vaeda_scores'] and adata.obs['vaeda_calls'] respectively.
For a more detailed and accessible example, see ./doc/vaeda_scanpy-pbmc3k-tutorial.ipynb, where we modified a tutorial from scanpy (https://scanpy-tutorials.readthedocs.io/en/latest/pbmc3k.html) to illustrate how to use vaeda for doublet annotation.