Installation

Install RustScenic as a Python package. Prebuilt packages are available for Linux, macOS and Windows; a GPU is not required.

Standard Install

pip install rustscenic

This documentation targets v0.5.0. The quickstart starts with network inference and activity scoring; the API map covers the correlation helpers and other analysis stages.

Supported Python versions: 3.10 to 3.13.

Supported platforms in the release workflow:

Platform Architecture
Linux x86_64, aarch64
macOS x86_64, aarch64
Windows x64

Core runtime dependencies are intentionally small:

  • numpy
  • pandas
  • pyarrow
  • scipy
  • anndata

Optional Extras

pip install "rustscenic[examples]"
pip install "rustscenic[validation]"
pip install "rustscenic[benchmarks]"

The reference extra includes pySCENIC ecosystem dependencies and is treated as informational because parts of that stack can fail on current Python packaging combinations. For strict reference comparisons, prefer the pinned Docker path under validation/reference/.

What This Avoids

The core RustScenic path does not require:

  • Java or Mallet for topic modelling.
  • dask or a distributed scheduler for local runs.
  • CUDA or GPU drivers.
  • Snakemake workflow setup.
  • A large pinned environment just to import the package.

Check The Install

python - <<'PY'
import rustscenic
import rustscenic.grn
import rustscenic.aucell
import rustscenic.cistarget
import rustscenic.topics

print(rustscenic.__version__)
print("core imports OK")
PY

Common Failure Modes

Symptom Likely cause Fix
TF list has zero overlap with expression matrix Gene-symbol mismatch, for example ENSEMBL IDs in var_names Use var["feature_name"] or pass matching TF names.
cisTarget NES is NaN Motif universe too small or zero variance across motifs Use a real motif ranking database with enough motifs.
pySCENIC reference install fails Upstream dependency conflict Use the pinned reference Docker path instead of relying on a fresh pip install.