Cellproject is a python package designed to assist with asymmetric scRNA-Seq data integration. By asymmetric integration we refer to various tasks also known as: cell projection or mapping, which attempt to transfer information from the reference landscape onto new data (‘target’) without modifying the reference. This includes: transferring cell annotation and fitting new data into the pre-existing PCA or UMAP coordinates.
Cellproject is based on the scanpy/anndata framework.
- correct/matching data scaling (often overlooked, to see why this matters have a look see notebook
cellproject_scaling.ipynb). - fitting new data into a pre-existing reference PC or UMAP space
- easy, nearest-neighbor-based cell annotation transfer
- nearest-neighbor regression method to fit new data into gene expression (even corrected or regressed values!) or PC space of the reference data
- a convenient python wrapper for Seurat data integration
- flexible design allowing asymmetric integration downstream of most (symmetric or asymmetric) batch correction tools
Basic cellproject functionality is covered in the notebook cellproject_overview.ipynb.
If you ever struggled to combined Smart-Seq2 and 10x data or need a practical walk-through have a look at the cellproject_use_case.ipynb notebook.
Python > 3.7 and pip are required. To install the package:
- Clone the repository:
git clone https://github.com/Iwo-K/cellproject
2 Install the package
pip install ./cellproject/
To use the python wrapper for Seurat batch correction, R installation is required with the following packages:
- anndata (CRAN)
- batchelor (Bioconductor)
- scater (Bioconductor)
- Seurat (CRAN)
30.03.2023 - fixed a bug in the run_seuratCCA function. Due to an unexpected behaviour of the ] operator in the Seurat package, the output may have been in the incorrect order if input data was not ordered by batch. run_seuratCCA uses now the anndata package for reordering and thus should work as expected.