Installation Guide¶
This document walks through the environment setup and installation steps required for CartLoader.
1. Dependencies¶
Before installing CartLoader, make sure the tools below are present on your system.
1.1 Required System Utilities¶
Confirm that these command-line programs are installed:
gzipsortbgziptabix
1.2 External Tools and Utilities¶
These packages support spatial data handling and file conversion. Several are bundled as git submodules.
Python & Related Tooling
python(verified with versions 3.10 and 3.13.1)parquet-tools
R & related Packages:
Rfrom CRAN(verified with versions 4.5.1)
External Tools (included as submodules)
punkst(the latest and more efficient implementation of FICTURE)spatulatippecanoemagickgo-pmtiles
Geospatial Utilities
Cloud & CLI Utilities
2. Setting Up the Environment using conda¶
We recommend isolating the project in a conda environment to avoid dependency conflicts.
2.1 Installing conda¶
If conda is not already available, download and install Miniconda or Anaconda.
Example installation of Miniconda3 on Linux:
1 2 3 4 5 | |
2.2 Creating an Environment¶
Set up a fresh environment for CartLoader:
1 2 | |
2.3 Install Core Dependencies¶
Install the core dependencies once the environment is active:
1 | |
3. Installing CartLoader¶
Clone the repository and install required Python and R packages:
1 2 3 4 5 6 7 8 9 10 11 | |
4. Initializing Submodules¶
1 2 | |
4.1 Installing spatula¶
Install spatula with its dependencies from the submodules directory:
1 2 3 4 5 6 7 8 9 10 11 | |
4.2 Installing punkst¶
Install the punkst toolkit to use FICTURE (Si et al., Nature Methods 2024).
Please follow the punkst installation guide.
What are FICTURE and punkst?
FICTURE is a segmentation-free method that infers latent spatial factors—coherent spatial patterns of gene activity—that correspond to underlying transcriptional programs or tissue structures. These factors can then be projected back to the pixel level. Although FICTURE is built on a Latent Dirichlet Allocation (LDA) framework by default, it is also compatible with clustering outputs from external tools like Seurat for pixel-level projection.
The punkst toolkit is a streamlined implementation of FICTURE, designed for improved computational efficiency and scalability while producing results equivalent to the original.
4.3 Installing tippecanoe¶
1 2 3 4 5 6 7 8 9 | |
4.4 Installing go-pmtiles¶
An easy way to install go-pmtiles is to download a release from the official website and decompress it. This provides a pmtiles binary ready for use.
Here is an example of its installation:
1 2 3 | |
4.5 Installing ImageMagick¶
Skip this step if ImageMagick was already installed via conda in Section 2.3.
1 2 3 4 | |
5. Verifying the Installation¶
Run the following command to verify that the package loads correctly:
1 | |