SI

single-cell-cellphonedb-communication-mapping

Analyzes single-cell data to map communication networks and generate publication-ready visualizations.

Install

mkdir -p .claude/skills/single-cell-cellphonedb-communication-mapping && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/3549" && unzip -o skill.zip -d .claude/skills/single-cell-cellphonedb-communication-mapping && rm skill.zip

Installs to .claude/skills/single-cell-cellphonedb-communication-mapping

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

CellPhoneDB v5 ligand-receptor analysis, CellChatViz plots, and the newer ccc_heatmap / ccc_network_plot / ccc_stat_plot communication visualizations in OmicVerse.
163 charsno explicit “when” trigger
Intermediate

Key capabilities

  • →Generate ccc_heatmap for communication intensity
  • →Build ccc_network_plot for cell-to-cell interaction
  • →Create ccc_stat_plot for quantifying signals
  • →Execute CellChatViz for customized visualization
  • →Validate AnnData communication column integrity

How it works

It pipes annotated expression data through OmicVerse wrappers that calculate ligand-receptor probabilities and apply standard visualization plotting APIs.

Inputs & outputs

You give it
Annotated AnnData object with cell labels
You get back
Communication maps, heatmaps, and network plots

When to use single-cell-cellphonedb-communication-mapping

  • →Quantify ligand-receptor communication
  • →Generate CellChat-style network plots
  • →Create communication heatmaps
  • →Visualize single-cell population interactions

About this skill

Single-cell CellPhoneDB communication mapping

Overview

Apply this skill when a user wants to quantify ligand-receptor communication between annotated single-cell populations and visualize the result with OmicVerse's CellPhoneDB plotting stack. This skill now covers both the original low-level ov.pl.CellChatViz workflow and the newer public plotting APIs:

  • ov.pl.ccc_heatmap(...)
  • ov.pl.ccc_network_plot(...)
  • ov.pl.ccc_stat_plot(...)

Use the ccc_* functions by default when the goal is to quickly generate publication-ready plots with a stable public API. Drop down to CellChatViz when the user needs method-level control or wants to combine several custom visualizations in one notebook.

Instructions

  1. Prepare the environment
    • Use an environment with omicverse>=0.2, scanpy, anndata, pandas, matplotlib, seaborn, and CellPhoneDB resources.
    • For the newer visualizations, also ensure these optional plotting dependencies are available when needed:
      • marsilea for heatmap and bubble matrix plotters.
      • mpl-chord-diagram for chord diagrams.
      • networkx for diffusion and network-style plots.
      • adjustText if the user wants improved automatic label repulsion.
    • Activate OmicVerse plotting defaults with ov.plot_set().
  2. Load and validate the annotated AnnData
    • Read the normalized expression matrix with ov.read(...).
    • Keep the communication grouping column clean, categorical, and aligned with the intended identities.
    • Recommended checks:
      celltype_key = "cell_labels"
      assert celltype_key in adata.obs.columns, f"{celltype_key} missing from adata.obs"
      adata.obs[celltype_key] = adata.obs[celltype_key].astype("category").cat.remove_unused_categories()
      assert not adata.obs[celltype_key].isna().any(), f"NaN values found in {celltype_key}"
      min_per_group = adata.obs[celltype_key].value_counts().min()
      if min_per_group < 10:
          print(f"WARNING: smallest group has {min_per_group} cells; sparse groups may destabilize permutations")
      
    • Confirm the matrix is log-normalized before running CellPhoneDB. Raw counts can distort permutation-based significance.
  3. Run CellPhoneDB through OmicVerse
    • Use ov.single.run_cellphonedb_v5(...) and persist the outputs:
      cpdb_results, adata_cpdb = ov.single.run_cellphonedb_v5(
          adata,
          cpdb_file_path="./cellphonedb.zip",
          celltype_key="cell_labels",
          min_cell_fraction=0.005,
          min_genes=200,
          min_cells=3,
          iterations=1000,
          threshold=0.1,
          pvalue=0.05,
          threads=10,
          output_dir="./cpdb_results",
          cleanup_temp=True,
      )
      
    • Save cpdb_results and adata_cpdb so downstream plotting can be repeated without rerunning permutations.
    • CellChat alternative (no R, no extra install beyond PyTorch): ov.single.run_cellchat reproduces R CellChat 2.2 (identical over-expressed genes/LR pairs, probabilities to ~1e-15) and stores adata.uns['cellchat_res'], which every ov.pl.ccc_* call reads directly:
      ov.single.run_cellchat(adata, groupby="cell_labels", species="human")  # log-normalised X
      comm = ov.single.format_cellchat_results(adata)                         # optional explicit comm AnnData
      df = ov.single.cellchat_subset_communication(adata, signaling=["TGFb"])  # significant LR table
      centr = ov.single.cellchat_centrality(adata)                             # sender/receiver/mediator/influencer
      
      CellChat scores are probabilities (~1e-3), not expression; pathway summaries pick scale-aware thresholds automatically. Groups with <= min_cells (default 10) cells are removed, as in R filterCommunication.
    • Unified entry point / several methods at once: ov.single.run_ccc runs any registered method (ov.single.available_ccc_methods(): cellchat, cellphonedb, liana, fastccc, secact, and the spatial cytosignal / commot) and always returns the same communication AnnData:
      comm = ov.single.run_ccc(adata, "cell_labels", method="cellchat")
      comm = ov.single.run_ccc(adata, "cell_labels", method=["cellchat", "cellphonedb", "liana"],
                               consensus="rra",                     # or "vote" (min_methods=...)
                               method_kwargs={"cellphonedb": {"cpdb_file_path": "./cellphonedb.zip"}})
      res = ov.single.compare_ccc(adata)                          # Jaccard / recall / rank correlations on shared-DB LR pairs
      ov.pl.ccc_compare(res, plot_type="upset")                   # also "jaccard", "recall", "score_corr", "top_lr", "cellpair"
      cp = ov.single.get_ccc(adata, method="cellphonedb")          # any single method's comm AnnData
      comm = ov.space.run_ccc(adata, "cell_type", method="cytosignal")  # spatial (needs obsm['spatial'])
      
      Interactions are matched across databases on (sender, receiver, ligand subunit genes, receptor subunit genes); CellPhoneDB receptor-first records are re-oriented ligand -> receptor. resource="shared" keeps only LR pairs present in every selected database. Per-method long tables live in adata.uns['ccc']['long'][method]; native results stay at cellchat_res / cpdb_results / liana_res.
  4. Prefer the new public plotting APIs for standard visualization requests
    • Use ov.pl.ccc_heatmap(...) for matrix-like plots:
      • plot_type="heatmap" for aggregated pathway-level communication.
      • plot_type="focused_heatmap" to highlight stronger interactions after thresholding weak entries.
      • plot_type="dot" or "bubble" for interaction-level summaries.
      • plot_type="pathway_bubble" for pathway-focused Marsilea bubble summaries.
      • plot_type="bubble_lr" for ligand-receptor-pair-specific bubble matrices.
      • plot_type="role_heatmap", "role_network", or "role_network_marsilea" for signaling role summaries.
      • plot_type="diff_heatmap" when comparing two communication AnnData objects.
    • Use ov.pl.ccc_network_plot(...) for graph-like plots:
      • plot_type="circle" or "circle_focused" for global communication networks.
      • plot_type="individual_outgoing" / "individual_incoming" for sender- or receiver-centric circle panels.
      • plot_type="individual" for a single pathway and optionally a selected L-R pair.
      • plot_type="chord" for cell-type-level pathway chords.
      • plot_type="gene_chord" for gene-level chord diagrams across pathway-specific ligand and receptor nodes.
      • plot_type="lr_chord" for specified ligand-receptor pairs.
      • plot_type="diffusion" for pathway similarity and diffusion-style network structure.
      • plot_type="diff_network" when comparing two communication objects.
      • plot_type="bipartite", "arrow", "sigmoid", or "embedding_network" for alternative layouts.
    • Use ov.pl.ccc_stat_plot(...) for statistics and summary panels:
      • plot_type="pathway_summary" to rank pathways by communication strength and significance.
      • plot_type="lr_contribution" to show the dominant ligand-receptor pairs within a pathway.
      • plot_type="scatter" or "role_scatter" to compare outgoing versus incoming signaling roles.
      • plot_type="role_network" or "role_network_marsilea" for matrix-style role summaries.
      • plot_type="sankey" for communication flow summaries.
  5. Use CellChatViz directly when the user needs method-level control
    • Create a stable palette mapping from cell labels:
      color_dict = dict(zip(
          adata.obs["cell_labels"].cat.categories,
          adata.uns["cell_labels_colors"]
      ))
      viz = ov.pl.CellChatViz(adata_cpdb, palette=color_dict)
      
    • Recommended direct workflow:
      • viz.compute_aggregated_network(...) then viz.netVisual_circle(...).
      • viz.compute_pathway_communication(...) then viz.get_significant_pathways_v2(...).
      • viz.netVisual_heatmap_marsilea(...) or viz.netVisual_heatmap_marsilea_focused(...).
      • viz.netVisual_bubble_marsilea(...) for pathway bubbles.
      • viz.netVisual_bubble_lr(...) for selected ligand-receptor pairs.
      • viz.netVisual_chord_cell(...), viz.netVisual_chord_gene(...), and viz.netVisual_chord_LR(...).
      • viz.netVisual_individual(...) for one pathway / one enriched pair.
      • viz.netAnalysis_computeCentrality() followed by role heatmap, scatter, and network plots.
      • viz.netAnalysis_contribution(pathway) for pathway-level pair contribution analysis.
      • viz.netVisual_diffusion(...) for pathway similarity structure.
  6. Highlight the new visualization capabilities clearly
    • The newer additions worth surfacing in answers are:
      • gene_chord: gene-level chord diagrams, not just cell-type-level chords.
      • bubble_lr: Marsilea bubble summaries centered on explicit ligand-receptor pairs.
      • focused_heatmap: thresholded pathway heatmaps that suppress weak interactions.
      • role_network_marsilea: richer role summaries with dendrograms, color bars, and importance bars.
      • diffusion: pathway similarity network based on communication patterns.
      • pathway_summary and lr_contribution: higher-level summary/statistical views for prioritization.
    • When the user says "new visualization", prioritize demonstrating one of these rather than only the legacy circle plot.
  7. Parameter tips for the newer plots
    • For pathway bubbles:
      • group_pathways=True groups by pathway rather than individual L-R pairs.
      • transpose=True is useful when too many cell-pair rows make labels unreadable.
      • add_violin=True can expose score distributions but makes figures denser.
    • For ligand-receptor bubbles:
      • show_all_pairs=True is useful when the user wants to compare a fixed panel of pairs even if some are weak or absent.

Content truncated.

When not to use it

  • →Datasets lacking annotated cell labels
  • →Studies without single-cell expression data

Prerequisites

omicverse>=0.2scanpyanndatamatplotlib

Limitations

  • →Depends on accuracy of cell population labels
  • →Requires specific plotting dependencies like marsilea or networkx

How it compares

It abstracts low-level plotting code into standardized OmicVerse APIs for quick publication-ready communication visuals.

Compared to similar skills

single-cell-cellphonedb-communication-mapping side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
single-cell-cellphonedb-communication-mapping (this skill)17moNo flagsIntermediate
llm-evaluation66moNo flagsAdvanced
evaluating-llms-harness310moReviewAdvanced
qutip411moReviewAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

llm-evaluation

wshobson

Implement comprehensive evaluation strategies for LLM applications using automated metrics, human feedback, and benchmarking. Use when testing LLM performance, measuring AI application quality, or establishing evaluation frameworks.

671

evaluating-llms-harness

davila7

Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag). Use when benchmarking model quality, comparing models, reporting academic results, or tracking training progress. Industry standard used by EleutherAI, HuggingFace, and major labs. Supports HuggingFace, vLLM, APIs.

337

qutip

davila7

Quantum mechanics simulations and analysis using QuTiP (Quantum Toolbox in Python). Use when working with quantum systems including: (1) quantum states (kets, bras, density matrices), (2) quantum operators and gates, (3) time evolution and dynamics (Schrödinger, master equations, Monte Carlo), (4) open quantum systems with dissipation, (5) quantum measurements and entanglement, (6) visualization (Bloch sphere, Wigner functions), (7) steady states and correlation functions, or (8) advanced methods (Floquet theory, HEOM, stochastic solvers). Handles both closed and open quantum systems across various domains including quantum optics, quantum computing, and condensed matter physics.

428

torchdrug

davila7

Graph-based drug discovery toolkit. Molecular property prediction (ADMET), protein modeling, knowledge graph reasoning, molecular generation, retrosynthesis, GNNs (GIN, GAT, SchNet), 40+ datasets, for PyTorch-based ML on molecules, proteins, and biomedical graphs.

326

string-database

davila7

Query STRING API for protein-protein interactions (59M proteins, 20B interactions). Network analysis, GO/KEGG enrichment, interaction discovery, 5000+ species, for systems biology.

217

transformer-lens-interpretability

davila7

Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.

215

Search skills

Search the agent skills registry