Palantir is an algorithm to align cells along differentiation trajectories. Palantir models differentiation as a stochastic process where stem cells differentiate to terminally differentiated cells by a series of steps through a low dimensional phenotypic manifold. Palantir effectively captures the continuity in cell states and the stochasticity in cell fate determination. Palantir has been designed to work with multidimensional single cell data from diverse technologies such as Mass cytometry and single cell RNA-seq.
-
Palantir has been implemented in Python3 and can be installed using:
pip install palantir
-
Palantir depends on a number of
python3
packages available on pypi and these dependencies are listed insetup.py
All the dependencies will be automatically installed using the above commands
-
To uninstall:
pip uninstall palantir
-
Palantir can also be used with Scanpy. It is fully integrated into Scanpy, and can be found under Scanpy's external modules (link)
A tutorial on Palantir usage and results visualization for single cell RNA-seq data can be found in this notebook: http://nbviewer.jupyter.org/github/dpeerlab/Palantir/blob/master/notebooks/Palantir_sample_notebook.ipynb
scanpy anndata
objects are available for download for the three replicates generated in the manuscript: Rep1, Rep2, Rep3
Each object has the following elements
.X
: Filtered, normalized and log transformed count matrix.raw
: Filtered raw count matrix.obsm['MAGIC_imputed_data']
: Imputed count matrix using MAGIC.obsm['tsne']
: tSNE maps presented in the manuscript generated using scaled diffusion components as inputs.obs['clusters']
: Clustering of cells.obs['palantir_pseudotime']
: Palantir pseudo-time ordering.obs['palantir_diff_potential']
: Palantir differentation potential.obsm['palantir_branch_probs']
: Palantir branch probabilities.uns['palantir_branch_probs_cell_types']
: Column names for branch probabilities.uns['ct_colors']
: Cell type colors used in the manuscript.uns['cluster_colors']
: Cluster colors used in the manuscript.varm['mast_diff_res_pval']
: MAST p-values for differentially expression in each cluster compared to others.varm['mast_diff_res_statistic']
: MAST statistic for differentially expression in each cluster compared to others.uns['mast_diff_res_columns']
: Column names for the differential expression results
Notebooks detailing the generation of results comparing Palantir to trajectory detection algorithms are available here
Use the snippet below to convert anndata
to Seurat
objects
library("SeuratDisk")
library("Seurat")
library("reticulate")
use_condaenv(<conda env>, required = T) # before, install "anndata" into <conda env>
anndata <- import('anndata')
#link to Anndata files
url_Rep1 <- "https://s3.amazonaws.com/dp-lab-data-public/palantir/human_cd34_bm_rep1.h5ad"
curl::curl_download(url_Rep1, basename(url_Rep1))
url_Rep2 <- "https://s3.amazonaws.com/dp-lab-data-public/palantir/human_cd34_bm_rep2.h5ad"
curl::curl_download(url_Rep2, basename(url_Rep2))
url_Rep3 <- "https://s3.amazonaws.com/dp-lab-data-public/palantir/human_cd34_bm_rep3.h5ad"
curl::curl_download(url_Rep3, basename(url_Rep3))
#H5AD files are compressed using the LZF filter.
#This filter is Python-specific, and cannot easily be used in R.
#To use this file with Seurat and SeuratDisk, you'll need to read it in Python and save it out using the gzip compression
#https://github.com/mojaveazure/seurat-disk/issues/7
adata_Rep1 = anndata$read("human_cd34_bm_rep1.h5ad")
adata_Rep2 = anndata$read("human_cd34_bm_rep2.h5ad")
adata_Rep3 = anndata$read("human_cd34_bm_rep3.h5ad")
adata_Rep1$write_h5ad("human_cd34_bm_rep1.gzip.h5ad", compression="gzip")
adata_Rep2$write_h5ad("human_cd34_bm_rep2.gzip.h5ad", compression="gzip")
adata_Rep3$write_h5ad("human_cd34_bm_rep3.gzip.h5ad", compression="gzip")
#convert gzip-compressed h5ad file to Seurat Object
Convert("human_cd34_bm_rep1.gzip.h5ad", dest = "h5seurat", overwrite = TRUE)
Convert("human_cd34_bm_rep2.gzip.h5ad", dest = "h5seurat", overwrite = TRUE)
Convert("human_cd34_bm_rep3.gzip.h5ad", dest = "h5seurat", overwrite = TRUE)
human_cd34_bm_Rep1 <- LoadH5Seurat("human_cd34_bm_rep1.gzip.h5seurat")
human_cd34_bm_Rep2 <- LoadH5Seurat("human_cd34_bm_rep2.gzip.h5seurat")
human_cd34_bm_Rep3 <- LoadH5Seurat("human_cd34_bm_rep3.gzip.h5seurat")
Thanks to Anne Ludwig from University Hospital Heidelberg for the tip!
Palantir manuscript is available from Nature Biotechnology. If you use Palantir for your work, please cite our paper.
@article{Palantir_2019,
title = {Characterization of cell fate probabilities in single-cell data with Palantir},
author = {Manu Setty and Vaidotas Kiseliovas and Jacob Levine and Adam Gayoso and Linas Mazutis and Dana Pe'er},
journal = {Nature Biotechnology},
year = {2019},
month = {march},
url = {https://doi.org/10.1038/s41587-019-0068-4},
doi = {10.1038/s41587-019-0068-4}
}
- Enable an AnnData-centric workflow for improved usability and interoperability with other single-cell analysis tools.
- Introduced new utility functions
palantir.utils.early_cell
To automate fining an early cell based on cell type and diffusion components.palantir.utils.find_terminal_states
To automate finding terminal cell states based on cell type and diffusion components.palantir.presults.select_branch_cells
To find cells associated to each branch based on fate probability.palantir.plot.plot_branch_selection
To inspect the cell to branch association.palantir.utils.run_local_variability
To compute local gene expression variability.palantir.utils.run_density
A wrapper for mellon.DensityEstimator.palantir.utils.run_density_evaluation
Evaluate computed density on a different dataset.palantir.utils.run_low_density_variability
. To aggregate local gene expression variability in low density.palantir.plot.plot_branch
. To plot branch-selected cells over pseudotime in arbitrary y-postion and coloring.palantir.plot.plot_trend
. To plot the gene trend ontop ofpalantir.plot.plot_branch
.
- Added input validation for better error handling and improved user experience.
- Expanded documentation within docstrings, providing additional clarity for users and developers.
- Updated tutorial notebook to reflect the new workflow, guiding users through the updated processes.
- Implemented gene trend computation using Mellon, providing more robust and efficient gene trend analysis.
- Enable annotation in
palantir.plot.highight_cells_on_umap
.
- Replaced PhenoGraph dependency with
scanpy.tl.leiden
for gene trend clustering. - Deprecated the
run_tsne
,determine_cell_clusters
, andplot_cell_clusters
functions. Use corresponding implementations from Scanpy, widely used single-cell analysis library and direct dependecy of Palantir. - Rename
palantir.plot.highight_cells_on_tsne
topalantir.plot.highight_cells_on_umap
- Depend on
anndata>=0.8.0
to avoid issues writing dataframes inad.obsm
.
- Addressed the issue of variability when reproducing results (issue#64), enhancing the reproducibility and reliability of Palantir.
- Replaced rpy2 with pyGAM for computing gene expression trends.
- Updated tutorial and plotting functions
- A fix to issue#41
- A fix to issue#42
- Revamped tutorial with support for Anndata and force directed layouts
- A fix related to issue#28. When identifying terminal states, duplicate values were generated instead of unique ones.