vaeda (variaitonal auto-encoder (vae) for doublet annotation (da)) is a computational tool for doublet annotation in single cell RNA-sequencing.
This is the code and results underlying our pre-print.
To install the python package vaeda, see the package's github for installation instructions.
Code to produce vaeda's results can be found in the dirrectory notebooks. Code for benchmarking results from scDblFinder, DoubletFinder, scds, and scrublet can be found in the dirrectory called R. Code for benchmarking solo can be found in the dirrectory scripts. Code to reproduce main figures can be found in the diffectory figures. The environment used to produce these results is available in vaeda_env.yaml. Lastly, doublet scores and calls calculated in this manuscript are available in results_benchmark.zip and results_downsample_cells.zip.
To reproduce figures from the manuscript, first, download the benchmarking datasets from zenodo. Convert the files from RDS to MTX format using the R script figures/rds2mtx.R. Then, unzip the zipped files. Then all figures can be reproduced by running the notebooks in the dirrectory figures.