scop

Overview

scop is an R/Seurat-centered single-cell omics package that delegates many methods to Python through reticulate. Treat setup as a two-layer install: keep the R package lean, then build and verify a dedicated Python environment.

Use this skill for local or Linux workstation setup, troubleshooting PrepareEnv(), checking Python module imports, and summarizing scop plot families. For generic Seurat/Scanpy preprocessing, use the single-cell or Scanpy skills instead.

Known-Good Local Baseline

Layer	Verified state
R package	scop 0.8.9 from mengxu98/scop; R 4.5.1
Seurat stack	Seurat 5.4.0, SeuratObject 5.4.0, reticulate 1.46.0
Python env	/home/h2048/miniconda3/envs/scop_env, Python 3.10.1
Safe shell prefix	env -u LD_LIBRARY_PATH -u PYTHONPATH PYTHONNOUSERSITE=1
Verification outputs	/home/h2048/temp/scop_verify_20260512
Plot catalog	/home/h2048/temp/scop_plot_catalog_20260512.tsv

Always run checks with the safe shell prefix in this workspace; user-site and base-conda pollution have caused false failures before.

Setup Recipe

1. Install the R package with core dependencies only. Avoid dependencies = TRUE unless the user explicitly needs every optional feature; it can pull fragile Suggests such as GitHub-only packages and system libraries.

pak::pkg_install(
  "mengxu98/scop",
  dependencies = c("Depends", "Imports", "LinkingTo")
)

2. Create or inspect the Python environment. scop::PrepareEnv() can bootstrap scop_env, but do not assume all default modules succeed on the first pass.

options(scop_envname = "scop_env")
scop::PrepareEnv(
  envname = "scop_env",
  conda = "/home/h2048/miniconda3/condabin/conda",
  pip_options = "-i https://pypi.tuna.tsinghua.edu.cn/simple"
)

scop::env_info is not exported in scop 0.8.9; use scop::ListEnv(), reticulate::py_config(), conda list, and direct Python imports instead.

3. Repair common Python/conda gaps. bedtools needs bioconda; conda-forge alone failed in the verified setup.

env -u LD_LIBRARY_PATH -u PYTHONPATH PYTHONNOUSERSITE=1 \
  /home/h2048/miniconda3/condabin/conda install --yes --name scop_env \
  -c conda-forge -c bioconda leidenalg=0.10.2 tbb=2022.2.0 \
  python-igraph=0.11.9 scvi-tools=1.2.1 bedtools

4. Preserve the conda PyTorch/scVI stack. Do not use uv pip install --force-reinstall for the full default module list; it attempted to pull huge CUDA/NVIDIA torch wheels and can destabilize the environment. Prefer normal pip install --no-user for missing packages.

5. Pin the known incompatibilities. With pandas==2.0.3 and jax/jaxlib==0.4.38, imports were fixed by:

env -u LD_LIBRARY_PATH -u PYTHONPATH PYTHONNOUSERSITE=1 \
  /home/h2048/miniconda3/envs/scop_env/bin/python -m pip install --no-user --no-deps \
  xarray==2023.12.0 optax==0.2.4

6. Install MultiMAP from codeload if GitHub clone fails. The package imports as MultiMAP, not lowercase multimap.

env -u LD_LIBRARY_PATH -u PYTHONPATH PYTHONNOUSERSITE=1 \
  /home/h2048/miniconda3/envs/scop_env/bin/python -m pip install --no-user \
  https://codeload.github.com/Teichlab/MultiMAP/zip/refs/heads/master

Validation Checklist

• R package loads: library(scop); packageVersion("scop").
• Core R dependencies load: Seurat, SeuratObject, reticulate, ggplot2, patchwork.
• Exported plot functions exist in asNamespace("scop").
• Python imports pass for: scanpy, anndata, scvi, scglue, scanorama, bbknn, celltypist, cellphonedb, magic, scrublet, doubletdetection, palantir, scvelo, cellrank, wot, phate, pacmap, trimap, MultiMAP, pybedtools, loompy.
• Isolated pip check reports no broken requirements.
• In this workspace, run /home/h2048/temp/verify_scop_setup_20260512.R to regenerate package/function TSVs and smoke-test figures.

Plot Function Families

The verified catalog has 39 non-interactive plotting entries plus the SCExplorer launcher row.

Family	Functions / modes
Embedding / projection	CellDimPlot, FeatureDimPlot, CellDimPlot3D, FeatureDimPlot3D, CellDensityPlot, ProjectionPlot
Heatmap / correlation	GroupHeatmap, FeatureHeatmap, CellCorHeatmap, FeatureCorPlot
Statistics / composition	CellStatPlot, FeatureStatPlot, ProportionTestPlot
Differential / enrichment	DEtestPlot, VolcanoPlot, DEtestManhattanPlot, DEtestRingPlot, EnrichmentPlot, GSEAPlot, GSVAPlot, MetabolismPlot, scTenifoldKnkPlot
Trajectory / dynamic	LineagePlot, BranchStreamPlot, DynamicPlot, DynamicHeatmap, PAGAPlot, PseudotimeProjectionPlot, VelocityPlot, CytoTRACEPlot
Spatial	SpatialDimPlot
Cell-cell communication	CCCHeatmap, CCCNetworkPlot, CCCStatPlot
Benchmark / genomic	BenchmarkPlot, LISIPlot, DimsEstimatePlot, CoverageTrackPlot, TACSPlot
Interactive	PrepareSCExplorer, RunSCExplorer

Example smoke-test outputs from the verified setup: CellDimPlot, FeatureDimPlot, CellStatPlot, and GroupHeatmap PNGs under /home/h2048/temp/scop_verify_20260512.

Common Failures

Symptom	Fix
R install fails while pulling optional GitHub/Suggests packages	Reinstall with dependencies = c("Depends", "Imports", "LinkingTo") first
bedtools not found during PrepareEnv() conda stage	Add -c bioconda to the conda install channels
scvi import fails through xarray / pandas API mismatch	Pin xarray==2023.12.0
optax import fails with current jax	Pin optax==0.2.4
MultiMAP GitHub clone fails due TLS/gnutls	Install the codeload master zip URL; import-check MultiMAP
Lowercase multimap import fails	This is expected; the installed module name is MultiMAP
Strange imports from base conda or ~/.local	Re-run with env -u LD_LIBRARY_PATH -u PYTHONPATH PYTHONNOUSERSITE=1
scop::env_info fails	It is not exported in 0.8.9; use ListEnv, py_config, and direct module probes