CAPITAL

Comparative pseudotime analysis of single-cell RNA-seq data

Last updated: 2020-11-13

We present CAPITAL, a method for comparing pseudotime trajectories with tree alignment whereby trajectories including branching can be compared without any knowledge of paths to be compared.

Installation

• CAPITAL (ver. 0.2.3) in Python

More user-friendly version (1.0.0) that can be used in an interactive development environment such as JupyterLab will be available soon.

Requirements

• Python>=3.8 (Miniconda is recommended)
• graphtools
• h5py<=2.10
• leidenalg
• magic-impute
• pygraphviz
• scanpy>=1.6
• scprep
• tslearn

Option

• jupyterlab

Install on Linux, Windows (WSL) and macOS

0. Create a new environment for CAPITAL if you want to keep your own Python environment built with conda:

$ conda create -n capital python=3.8

Then, activate the environment:

$ conda activate capital

1. Install Scanpy and others via conda:

$ conda install -c bioconda scanpy
$ conda install graphtools leidenalg pygraphviz scprep tslearn

Note: the above command may automatically install h5py>=3.0, which will cause the problem of having bytes labels in H5AD data. To circumvent this, try:

$ conda install h5py=2

2. Install MAGIC via pip:

$ pip install --no-deps magic-impute

3. Download the tarball, and type the followings in your terminal:

$ tar zxf capital-0.2.3.tar.gz
$ cd capital-0.2.3

Pipeline

CAPITAL consists of three python codes. The standard use of CAPITAL is illustrated with the following steps:

Step 1: run pre_capital.py to preprocess raw data of scRNA-seq gene expression for each experiment

$ ./pre_capital.py [option]* <data1>
$ ./pre_capital.py [option]* <data2>

Usage: pre_capital.py

positional arguments:
  data <STR>            path to the raw [H5AD|CSV|TXT] file or 10x-MTX directory of scRNA-seq gene expression profile

optional arguments:
  -h, --help            show this help message and exit
  -t, --transpose       transpose the data [off]
  --min-genes <INT>     keep cells with at least <INT> genes expressed [200]
  --min-cells <INT>     keep genes that are expressed in at least <INT> cells [3]
  --magic               impute gene expression data with MAGIC [off]
  -n <INT>, --top-n-genes <INT>
                        number of highly variable genes to keep [2000]
  -p <INT>, --n-pcs <INT>
                        number of principal components for computing a k-nearest neighbor graph and a tree [50]
  -k <INT>, --neighbors <INT>
                        compute an <INT>-nearest neighbor graph [10]
  --no-save             results are not saved [off: saved in ./processed_data]
  -f <STR>, --filename <STR>
                        save data as <filename>.h5ad and umap_<filename>.pdf
  --save-fig            save a UMAP PDF figure in ./figures [off]

Step 2: run capital.py to predict pseudotime trajectories for two preprocessed data, and compute a trajectory alignment along with associated dynamic time warping of each aligned path

$ ./capital.py [option]* <data1> <data2> <root1> <root2> <genes>

Usage: capital.py

positional arguments:
  data1 <STR>           path to the preprocessed expression H5AD data for experiment 1 generated with pre_capital.py
  data2 <STR>           path to the preprocessed expression H5AD data for experiment 2 generated with pre_capital.py
  root1 <STR>           root cluster in data1
  root2 <STR>           root cluster in data2
  genes <STR>           path to the file that contains gene names to be analyzed (one gene per line)

optional arguments:
  -h, --help            show this help message and exit
  -M {euclid,gauss,paga}, --method {euclid,gauss,paga}
                        method for calculating a tree [euclid]
  -d {pca,diffmap}, --dimension {pca,diffmap}
                        dimension metric for calculating a tree [pca]
  -c <FLOAT>, --gapcost <FLOAT>
                        gap cost for calculating a tree alignment [1.0]
  -m <INT>, --n-genes1 <INT>
                        number of highly variable genes in data1 [2000]
  -n <INT>, --n-genes2 <INT>
                        number of highly variable genes in data2 [2000]
  -t, --tune            tuning mode, which affects naming of the result directory and never saves H5AD data [off]
  --no-prune            disable pruning of space nodes on edges of each alignment path for dynamic time warping [off]

Step 3: run draw_capital.py to draw figures on dynamic time warping and/or expression dynamics for a gene in aligned_data created in Step 2

$ ./draw_capital.py [option]* <alignment> <genes>

Usage: draw_capital.py

positional arguments:
  alignment <STR>     path to the directory for aligned data generated with capital.py (e.g. ./aligned_data/data1_data2/alignment000/)
  genes <STR>         path to the file that contains gene names to be analyzed (one gene per line)

optional arguments:
  -h, --help          show this help message and exit
  --dtw <STR>         path to the directory for a figure on dynamic time warping. Unless specified, the figure will be saved in each gene file in    
                      the alignment directory
  --dyn <STR>         path to the directory for a figure on gene expression dynamics. Unless specified, the figure will be saved in each gene file   
                      in the alignment directory
  --data1-name <STR>  data1 name on the plot
  --data2-name <STR>  data2 name on the plot

Reference

Reiichi Sugihara, Yuki Kato, Tomoya Mori and Yukio Kawahara, Alignment of time-course single-cell RNA-seq data with CAPITAL, Preprint bioRxiv at https://doi.org/10.1101/859751, 2019.

---

If you have any questions, please contact Yuki Kato
Graduate School of Medicine, Osaka University, Japan