# 11. Session info

The versions this tutorial was run with.

## Python

```python
import session_info2

session_info2.session_info(dependencies=True)
```

`scanpy` prints the same report via `sc.logging.print_header()`.

| Package | Version |
|---------|---------|
| Python | 3.12.3 |
| scanpy | 1.12 |
| anndata | 0.12.19 |
| numpy | 2.4.6 |
| scipy | 1.18.0 |
| pandas | 2.2.2 |
| scikit-learn | 1.9.0 |
| statsmodels | 0.14.6 |
| matplotlib | 3.11.1 |
| seaborn | 0.13.2 |
| adjustText | 1.4.0 |
| networkx | 3.6.1 |
| python-igraph | 1.0.0 |
| leidenalg | 0.12.0 |
| liana | 1.8.1 |
| decoupler | 2.2.0 |
| intratalkerpy | `74fa5bb` |
| pycrosstalker | `a057b01` |
| deltacorrpy | `bb249bd` |

The last three are installed from git, so they are pinned by commit rather than by version:

```bash
pip install git+https://github.com/CostaLab/IntraTalkerpy.git@74fa5bb
pip install git+https://github.com/CostaLab/pyCrossTalkeR.git@a057b01
pip install git+https://github.com/vckraemer/deltacorrpy.git@bb249bd
```

`deltacorrpy` is pulled in by `intratalkerpy`; it provides the partial correlation kernel that
`project_perturbation_in_embedding` ([step 9](09_perturbation.md)) runs on.

## R

The R half — the IntraTalker/CrossTalkeR integration of [step 4](04_intratalker_crosstalker.md)
and the Sankey figures of [step 5](05_sankey_plots.md) — was run with:

```r
sessionInfo()
```

| Package | Version |
|---------|---------|
| R | 4.6.0 (x86_64-pc-linux-gnu) |
| Seurat | 5.5.0 |
| SeuratObject | 5.4.0 |
| CrossTalkeR | 2.0.0 |
| LR2TF | 0.1.0 |
| IntraTalker | 0.1.0 |
| Matrix | 1.7.5 |
| igraph | 2.3.1 |
| ggraph | 2.2.2 |
| ggplot2 | 4.0.3 |
| ggalluvial | 0.12.6 |
| dplyr | 1.2.1 |
| tidyr | 1.3.2 |
| scales | 1.4.0 |
| colorspace | 2.1.2 |
| colorBlindness | 0.1.9 |
| oce | 1.8.3 |

## Randomness

Everything that draws random numbers is seeded in the page where it runs:

| Step | Seed | Fixes |
|------|------|-------|
| [Communication communities](06_communities.md) | `seed=12` | the Leiden partition and both spring layouts |
| [KO simulations](09_perturbation.md) | `np.random.seed(15037)`, reset **inside** the receptor loop | the neighbour subsampling of the embedding projection |

The MSC trajectory comes ready to use with the object of [step 8](08_trajectory.md), so the
diffusion map and the pseudotime are the same on every run. The R half is deterministic given
its input CSVs.