Package
Meta

py-cellcall

Python port of cellcall: inference of intercellular and intracellular networks from single-cell transcriptomics.

Features

• Pure Python — no R dependency
• Fast — 4.8x speedup over R (379s vs 1768s on 35K genes x 366 cells)
• High parity — Pearson >= 0.998 for expr_mean and expr_l_r_log2_scale
• Compatible API with R cellcall

Install

pip install -e .

Dependencies

numpy>=1.22, pandas>=1.4, scipy>=1.8, statsmodels>=0.13

Quick Start

import pandas as pd
from cellcall.connect_profile import create_nichcon_object, trans_commu_profile

# Load count matrix (genes x cells)
# Column names must encode cell type: "BARCODE_CELLTYPE"
data = pd.read_csv("counts.csv", index_col=0)

# Create CellInter object
obj = create_nichcon_object(
    data=data,
    source="UMI",        # "UMI", "fullLength", "TPM", "CPM"
    org="Homo sapiens",  # or "Mus musculus"
)

# Run communication analysis
obj = trans_commu_profile(
    obj,
    pValueCor=0.05,
    CorValue=0.1,
    topTargetCor=1,
    method="weighted",   # "weighted", "max", "mean"
    use_type="median",
    probs=0.75,
)

# Access results
scores = obj.data["expr_l_r_log2_scale"]  # Final scaled scores
regulons = obj.data["regulons_matrix"]     # TF regulon scores
means = obj.data["expr_mean"]             # Mean expression per cell type

Function Map (R -> Python)

Output Slots

Visualization

Sankey Diagram (Ligand → Receptor → TF)

GSEA Enrichment Plot

L-R Communication & Regulon Heatmap

Citation

If you use cellcall, please cite:

Liu, T. et al. cellcall: integrated analysis of ligand-receptor and transcription factor activities from single-cell transcriptomics. (2021)