Installation

CyteType has two clients:

cytetype, the Python package, for AnnData and Scanpy workflows;
CyteTypeR, the R package, for Seurat workflows.

Both prepare your data on your own computer and send it to the CyteType service at https://cytetype.nygen.io, where the annotation runs. Install the client for the environment you analyze your data in. You can install both; they share the same sign-in.

Requirements¶

	Python	R
Language version	Python 3.12 or newer	R 4.2 or newer (see R notes)
Package source	PyPI	GitHub, with dependencies from CRAN and Bioconductor
Single-cell toolkit	AnnData, installed with `cytetype`	Seurat, installed with `CyteTypeR`

You also need internet access to https://cytetype.nygen.io and a CyteType account, which needs an institutional email address (see Sign in and API keys).

Install the client¶

Python

With pip, in the virtual environment or conda environment you use for your analysis:

pip install cytetype

With uv:

# Add CyteType to a uv project
uv add cytetype

# Or install it into the active virtual environment
uv pip install cytetype

The package also installs the cytetype command, which you use to sign in.

With devtools, as in the CyteTypeR README:

install.packages("devtools")
devtools::install_github("NygenAnalytics/CyteTypeR")

Or with pak:

install.packages("pak")
pak::pak("NygenAnalytics/CyteTypeR")

Both installers also install the packages that CyteTypeR requires, including Seurat, duckdb and the Bioconductor package rhdf5 (see R notes).

Python notes¶

Check your Python version first with python --version. pip only offers releases that support the Python you run it with. On Python 3.11, pip install cytetype installs release 0.12.0, the last one for 3.11, without any error. On older versions it finds no release at all.
Prefer a dedicated environment. cytetype requires narrow version ranges of some common packages (anndata 0.12.x, pydantic 2.12.x and requests 2.32.x), so installing it into an existing environment can upgrade or downgrade them.

R notes¶

R 4.2 or newer. CyteTypeR requires it, and so does the duckdb package it uses. Check your version with R.version.string, and update R first if it is older.
duckdb writes the cell metadata file that is uploaded with each job, and both installers add it for you. On Linux, CRAN installs duckdb from source, which can take a while.
rhdf5 from Bioconductor. CyteTypeR writes the expression file with rhdf5, a Bioconductor package, and both installers add it for you. pak installs it from the Bioconductor repositories, which provide binary packages for Windows and macOS. devtools looks it up on Bioconductor’s Git server instead, which needs Git (or the git2r package) on your computer, and when it installs rhdf5 it builds it from source, which needs a compiler (Rtools on Windows, the Xcode Command Line Tools on macOS). If you do not have these tools, use pak. If rhdf5 fails to build with devtools, install it from Bioconductor first and tell devtools to keep it:
```
install.packages("BiocManager")
BiocManager::install("rhdf5")

# Keep the installed rhdf5 instead of rebuilding it
devtools::install_github("NygenAnalytics/CyteTypeR", upgrade = "never")
```
Seurat and the other dependencies, such as SeuratObject and withr, are installed from CRAN when they are missing, so the first installation can take some time.

Optional packages that CyteTypeR uses when they are installed (the installers do not add them):

Package	What it is for
`furrr` and `progressr`	With furrr, CyteTypeR uploads large files in parallel chunks (up to `upload_max_workers`, 6 by default); with the default `show_progress = TRUE` it also needs progressr. Without them, chunks are uploaded one at a time.

Install the packages for the quickstart¶

The quickstarts use the standard Scanpy and Seurat workflows to cluster the example data before annotating it.

Python

Scanpy is not installed with cytetype. Install it with the leiden extra:

pip install "scanpy[leiden]"

# or, with uv
uv add "scanpy[leiden]"

The extra installs igraph and leidenalg, the two Leiden implementations Scanpy can use; the quickstart’s sc.tl.leiden(flavor="igraph") needs igraph. CyteType’s plotting helper, cytetype.marker_dotplot(), also needs Scanpy.

Check the installation¶

Python

cytetype --version
python -c "import cytetype; print(cytetype.__version__)"

Both print the installed version, for example cytetype 0.21.0 and 0.21.0.

If your shell cannot find the cytetype command, the environment’s scripts folder is not on your PATH. Activate the environment, or run the command through Python instead, for example python -m cytetype.cli --version.

Update¶

Python

pip install -U cytetype

# or, with uv in a virtual environment
uv pip install -U cytetype

# or, with uv in a project
uv lock --upgrade-package cytetype
uv sync

When Python imports cytetype, and when you run the cytetype command, the package asks PyPI for the latest version, waiting at most 2 seconds. If a newer release exists, it prints a warning such as:

⚠️  A newer CyteType version is available on PyPI (<installed> -> <latest>). Run `pip install -U cytetype` to update.

The check stays silent when you are offline or PyPI cannot be reached.

devtools::install_github("NygenAnalytics/CyteTypeR", force = TRUE)

# or, with pak
pak::pak("NygenAnalytics/CyteTypeR")

force = TRUE reinstalls the package even if devtools thinks it is up to date.

When you load the package with library(CyteTypeR), it asks GitHub for the latest CyteTypeR release, waiting at most 2 seconds. If a newer release exists, it prints a startup message such as:

A newer version of CyteTypeR is available (<installed> -> <latest>). Run devtools::install_github('NygenAnalytics/CyteTypeR') to update.

The check stays silent when you are offline or GitHub cannot be reached.

After updating, restart your Python kernel or R session so that the new version is loaded.

Optional: the `cytetyper` command for R¶

Python users get the cytetype terminal command with the package. R users can install an equivalent command, cytetyper, which runs the CyteTypeR sign-in functions from a terminal:

library(CyteTypeR)
InstallCyteTypeRCli()

The function prints and returns the path of the installed command. cytetyper has the same commands as the Python cytetype command:

cytetyper setup           # sign in with your browser (also: cytetyper get-key)
cytetyper login           # save an existing API key
cytetyper logout          # remove the saved key from this computer
cytetyper dashboard       # open the dashboard
cytetyper view <JOB_ID>   # open the report of a job
cytetyper --version

cytetyper only signs you in and opens pages. Annotation jobs always run from an R session with CyteTypeR(). You do not need the command at all: the R functions SetupCyteTypeR(), LoginCyteTypeR() and LogoutCyteTypeR() do the same things.

Before you rely on it, check these points:

It must stay next to Rscript. The command runs the Rscript in its own folder, so InstallCyteTypeRCli() copies it into the folder of the first Rscript on your R session’s PATH, after following symbolic links (with Homebrew or the CRAN installer for macOS, that is R’s own program folder). Keep the default destination; a copy in a folder without Rscript does not work. If you have several R versions, check that this Rscript belongs to the R where CyteTypeR is installed. If the function stops with Could not locate Rscript on PATH, add R’s bin folder to PATH and try again.
That folder must be on your PATH. Otherwise a new terminal cannot find cytetyper. With the R installers from CRAN for macOS and Windows, R’s own program folder is usually not on PATH: add it, or run the command with the full path that InstallCyteTypeRCli() printed. On Windows the installed file is cytetyper.cmd.
A system-wide R may need administrator rights. If you cannot write to R’s program folder, the function stops with Could not install cytetyper in <folder>. Check that the directory is writable. Ask your administrator, or use the R functions instead.
It ignores your R startup files. The command starts R with --vanilla, which skips .Rprofile and .Renviron. If CyteTypeR is installed in a project library (for example with renv) or in a library that only those files point to, the command fails with Error: Unable to load CyteTypeR.
Reinstall it for a new R version. A new R installation has its own program folder; run InstallCyteTypeRCli() again from the new version. Use overwrite = TRUE to replace an existing copy.

Next step¶

Sign in once on each computer you run CyteType from: Sign in and API keys. If the installation fails, see Troubleshooting.

Requirements¶

Install the client¶

Python notes¶

R notes¶

Install the packages for the quickstart¶

Check the installation¶

Update¶

Optional: the cytetyper command for R¶

Next step¶

Optional: the `cytetyper` command for R¶