Installation

Release binaries

Linux x86-64:

curl -fLO https://github.com/kharchenkolab/Baysor/releases/download/cpp-0.9.0/baysor-0.9.0-linux-x86_64.tar.gz
tar -xzf baysor-0.9.0-linux-x86_64.tar.gz
./baysor-0.9.0-linux-x86_64/bin/baysor run -m 30 -s 8 molecules.csv

30 and 8 are example parameters; see Cell segmentation before choosing them for your data.

Download other platforms and SHA256SUMS from the cpp-0.9.0 release:

Archive	Requirements
baysor-0.9.0-linux-x86_64.tar.gz	Linux x86-64, glibc 2.28 or newer (e.g. Debian 10+, RHEL 8+)
baysor-0.9.0-macos-arm64.tar.gz	macOS 12 or newer, Apple Silicon
baysor-0.9.0-windows-x86_64.zip	64-bit Windows 10 or newer

Extract the archive and use bin/baysor (bin/baysor.exe on Windows) inside the extracted directory. Keep the Windows DLLs beside the executable. The binaries use generic CPU baselines; AVX2 is not required.

Optionally verify the Linux archive before extracting it:

curl -fL -o SHA256SUMS \
  https://github.com/kharchenkolab/Baysor/releases/download/cpp-0.9.0/SHA256SUMS
sha256sum --check --ignore-missing SHA256SUMS

Docker

From the directory containing molecules.csv:

docker run --rm -v "$PWD:/data" ghcr.io/kharchenkolab/baysor:0.9.0 run -m 30 -s 8 molecules.csv

The release image is Linux x86-64 and uses baysor as its entrypoint: pass run, preview or segfree directly after the image name. /data is its working directory, so paths are relative to the mounted directory. The image runs as UID/GID 1000; if your user has another ID, add --user "$(id -u):$(id -g)" so the results are writable.

Version tags (such as 0.9.0) pin a release; latest tracks the newest stable release. The older Docker Hub images vpetukhov/baysor (v0.4–v0.7.1) contain the Julia implementation.

Building from source

Use this only if a release binary does not suit your platform or you need to modify Baysor. On Linux / macOS, the simplest build uses a private conda-forge environment bootstrapped by configure.sh; no root access is needed:

git clone -b cpp-0.9.0 https://github.com/kharchenkolab/Baysor.git
cd Baysor
./configure.sh --deps=conda --install
./install/bin/baysor --version

The default dependency directory is .deps; the binary is installed in install/bin. Other dependency modes are system (packages already installed), vcpkg (requires VCPKG_ROOT) and auto (default: system if the build tools and Arrow/Parquet are available, otherwise conda).

Using system packages

A source build needs a C++17 compiler, CMake ≥ 3.20, a build tool, Eigen ≥ 3.3, spdlog, CGAL, Arrow/Parquet, HDF5, nlohmann_json, libtiff and zlib. CMake fetches the header-only UMAP dependencies automatically.

On Ubuntu 24.04, first enable the Apache Arrow apt repository:

sudo apt-get update
sudo apt-get install -y --no-install-recommends ca-certificates lsb-release wget
wget https://packages.apache.org/artifactory/arrow/$(lsb_release --id --short | tr 'A-Z' 'a-z')/apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
sudo apt-get install -y --no-install-recommends ./apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
rm ./apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
  build-essential cmake ninja-build pkg-config git \
  libeigen3-dev libspdlog-dev libcgal-dev libarrow-dev libparquet-dev \
  libhdf5-dev nlohmann-json3-dev libtiff-dev zlib1g-dev
./configure.sh --deps=system --install

On macOS with Homebrew:

brew install cmake ninja pkg-config eigen spdlog cgal apache-arrow hdf5 nlohmann-json libtiff
./configure.sh --deps=system --install

For dependencies in a non-standard location, pass their prefix to CMake:

./configure.sh --deps=system -- -DCMAKE_PREFIX_PATH=/path/to/prefix

The repository also provides a source-build Dockerfile: docker build -t baysor . from the repository root. See ./configure.sh --help and Development for build and test options. If configuration fails, the error lists the missing dependency; include the configure log when opening an issue.

Legacy Julia implementation

For Baysor.jl v0.7.1, follow the migration page and archived installation guide. Do not use an unpinned Julia package install: the default branch is now C++.