Skip to content

METHODS.md

Latest commit

 

History

History
239 lines (174 loc) · 5.83 KB

METHODS.md

File metadata and controls

239 lines (174 loc) · 5.83 KB
Raw

SpatialClaw Methods Guide

This guide summarizes the spatial-only skill set and canonical CLI names.

Preprocessing

Canonical skill: spatial-preprocessing

python spatialclaw.py run spatial-preprocessing \
  --input data.h5ad \
  --output output/spatial_preprocessing

Key parameters include --data-type, --species, --min-genes, --min-cells, --max-mt-pct, --n-top-hvg, --n-pcs, --n-neighbors, and --leiden-resolution.

Spatial Domain Identification

Canonical skill: spatial-domain-identification

Methods include leiden, louvain, spagcn, stagate, graphst, and banksy when dependencies are installed.

python spatialclaw.py run spatial-domain-identification \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_domain_identification \
  --method leiden --resolution 0.8

Cell Type Annotation

Canonical skill: spatial-cell-annotation

Methods include marker-based annotation, Tangram, scANVI, and CellAssign.

python spatialclaw.py run spatial-cell-annotation \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_annotation \
  --method marker_based --species human

Deconvolution

Canonical skill: spatial-deconvolution

Methods include Tangram, Stereoscope, and GraphST.

python spatialclaw.py run spatial-deconvolution \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_deconvolution \
  --method tangram \
  --reference ref.h5ad \
  --cell-type-key cell_type

Spatial Statistics

Canonical skill: spatial-statistics

Analysis types include neighborhood enrichment, Moran's I, Geary's C, local Moran, Getis-Ord, bivariate Moran, Ripley, co-occurrence, network properties, and spatial centrality.

python spatialclaw.py run spatial-statistics \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_statistics \
  --analysis-type moran --n-top-genes 20

Spatially Variable Genes

Canonical skill: spatial-svg-detection

Methods include Moran's I, SpatialDE, and FlashS.

python spatialclaw.py run spatial-svg-detection \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_svg \
  --method morans --n-top-genes 50

Differential Expression

Canonical skill: spatial-de

python spatialclaw.py run spatial-de \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_de \
  --method wilcoxon --groupby leiden --n-top-genes 20

Condition Comparison

Canonical skill: spatial-condition-comparison

python spatialclaw.py run spatial-condition-comparison \
  --input data.h5ad \
  --output output/spatial_condition_comparison \
  --condition-key treatment \
  --sample-key sample_id \
  --reference-condition control

Cell Communication

Canonical skill: spatial-cell-communication

Methods include built-in ligand-receptor scoring, LIANA+, CellPhoneDB, and FastCCC when optional Python dependencies are available.

python spatialclaw.py run spatial-cell-communication \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_cell_communication \
  --method liana --cell-type-key leiden --species human

Velocity And Trajectory

Canonical skills: spatial-velocity, spatial-trajectory

python spatialclaw.py run spatial-velocity \
  --input data.h5ad \
  --output output/spatial_velocity \
  --method stochastic

python spatialclaw.py run spatial-trajectory \
  --input output/spatial_velocity/processed.h5ad \
  --output output/spatial_trajectory \
  --method dpt

Enrichment

Canonical skill: spatial-enrichment

python spatialclaw.py run spatial-enrichment \
  --input output/spatial_de/processed.h5ad \
  --output output/spatial_enrichment \
  --analysis-type go \
  --groupby leiden \
  --organism human

CNV

Canonical skill: spatial-cnv

python spatialclaw.py run spatial-cnv \
  --input output/spatial_preprocessing/processed.h5ad \
  --output output/spatial_cnv \
  --method infercnvpy --reference-key cell_type

Multi-Sample Integration

Canonical skill: spatial-integration

python spatialclaw.py run spatial-integration \
  --input combined.h5ad \
  --output output/spatial_integration \
  --method harmony --batch-key batch

Spatial Registration

Canonical skill: spatial-registration

python spatialclaw.py run spatial-registration \
  --input combined.h5ad \
  --output output/spatial_registration \
  --method paste --reference-slice slice1.h5ad

Spatial Modality Integration

Canonical skill: spatial-modality-integrate

This skill accepts directory-based spatial platform inputs for DeepST or PearlST.

python spatialclaw.py run spatial-modality-integrate \
  --input /path/to/DLPFC/151673 \
  --output output/spatial_modality \
  --method deepst --n-domains 7 --use-gpu

Spatial Paired-Modality Integration

Canonical skill: spatial-omics-integrate

This is a spatial skill for paired spatial modalities such as RNA plus protein or RNA plus ATAC.

python spatialclaw.py run spatial-omics-integrate \
  --input rna.h5ad \
  --omics2 protein.h5ad \
  --output output/spatial_omics \
  --method spatialglue \
  --omics1-type rna \
  --omics2-type protein

Current Registry

Use python spatialclaw.py list for the active registry. Public skill documentation and routing are limited to registered CLI-capable spatial skills. Internal helpers under skills/spatial/_lib/ are implementation modules, not standalone skills.

Orchestrator

Canonical skill: spatial-orchestrator

python spatialclaw.py run spatial-orchestrator \
  --query "find spatially variable genes" \
  --output output/spatial_route

python spatialclaw.py run spatial-orchestrator \
  --pipeline standard \
  --input data.h5ad \
  --output output/spatial_pipeline