# 6. Communication communities

**Language:** Python · **Input:** `edges_directed_KO_WT_filtered.csv`

In this step, we will group the cell types into communication communities from the
differential cell–cell network. The CrossTalkeR2 community-detection procedure runs on the
edge list exported in [step 4](04_intratalker_crosstalker.md) and nothing else — no
expression data is required.

## Detection of communication communities

Currently, the function is still situated on the dev branch of the pyCrossTalkeR repository. Thus we install it to perform the clustering and generate the communication community plot.

```bash
pip install git+https://github.com/CostaLab/pyCrossTalkeR.git@dev
```
And then we can generate and plot the communication community layout:
```python
from pathlib import Path
import pandas as pd
import pycrosstalker as pyct

BASE       = Path(Path("/path/to/Il1rn_KO"))
CTR_OUT    = BASE / "crosstalker"

df = pd.read_csv(CTR_OUT / "edges_directed_KO_WT_filtered.csv")

pyct.plots.cci_community_layout(
    df,
    method="leiden",
    res=1.3,
    compression=0.3,
    seed=12,
    spring_k=0.1,
    spring_iterations_community=1000,
    spring_iterations_nodes=500,
    edge_width_scale=2.5,
    edge_alpha_intra=1,
    edge_alpha_inter=1,
    node_size=300,
    arrow_size=20,
    figsize=[10, 10])
```

:::{figure} images/step6_cci_community_KO_WT.png
:alt: Cell-cell communication communities (Leiden)
:width: 80%
:::

At `res=1.3`, four communities:

| Community | Members | Direction in KO |
|-----------|---------|-----------------|
| Homeostasis (stem/progenitor) | HSC, MPPs, CMP | **decreased** communication |
| Myeloid–Vascular | monocytes, preDC, pericytes, sinusoidal | increased |
| Neutrophil | G0–G4 | increased |
| Stromal | MSCs, fibroblasts, arteriolar | increased |

MSCs carry the highest signal in the network, and stromal–stromal interactions are
predominantly upregulated.


All following steps are for running the perturbation module of IntraTalker.

Next: [Receptome construction](07_receptome.md)