---
title: "Python Integration with anndataR"
package: anndataR
output: BiocStyle::html_document
vignette: >
  %\VignetteIndexEntry{Python integration with anndataR}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
reticulate::py_require("mudata")
reticulate::py_require("scanpy")

# Check if required Python packages are available
python_available <- reticulate::py_module_available("scanpy") &&
  reticulate::py_module_available("mudata")

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  # Set eval based on Python package availability
  eval = python_available
)
```

# Introduction

`r Biocpkg("anndataR")` works with Python `AnnData` objects through `r CRANpkg("reticulate")`.
You can load Python objects, apply Python functions to them, and convert them to `Seurat` or `SingleCellExperiment` objects.

```{r python_unavailable, eval=!python_available}
message(
  "Python packages scanpy and mudata are required to run this vignette. Code chunks will not be evaluated."
)
```

## Prerequisites

This vignette requires Python with the _scanpy_ and _mudata_ packages installed.
If these are not available, the code chunks will not be evaluated but the examples remain visible.

# Basic Integration with _scanpy_

Install required Python packages if needed:

```{r python_setup, eval=FALSE}
reticulate::py_require("scanpy")
```

```{r load_libraries}
library(anndataR)
library(reticulate)
sc <- import("scanpy")
```

Load a dataset directly from _scanpy_:

```{r load_python_data}
adata <- sc$datasets$pbmc3k_processed()
adata
```

Apply _scanpy_ functions directly:

```{r apply_python_functions}
sc$pp$filter_cells(adata, min_genes = 200L)
sc$pp$normalize_total(adata, target_sum = 1e4)
sc$pp$log1p(adata)
```

# Conversion to R objects

Convert to `SingleCellExperiment` (see `vignette("usage_singlecellexperiment")`):

```{r convert_to_sce}
sce_obj <- adata$as_SingleCellExperiment()
sce_obj
```

Convert to `Seurat` (see `vignette("usage_seurat")`):

```{r convert_to_seurat}
seurat_obj <- adata$as_Seurat()
seurat_obj
```

# Multi-modal data with _mudata_

Install required Python packages if needed:

```{r mudata_setup, eval=FALSE}
reticulate::py_install("mudata")
```

```{r load_mudata}
md <- import("mudata")
```

Load a `MuData` object from file:

```{r load_mudata_example, eval = python_available && requireNamespace("BiocFileCache", quietly = TRUE)}
cache <- BiocFileCache::BiocFileCache(ask = FALSE)
h5mu_file <- BiocFileCache::bfcrpath(
  cache,
  "https://github.com/gtca/h5xx-datasets/raw/b1177ac8877c89d8bb355b072164384b4e9cc81d/datasets/minipbcite.h5mu"
)

mdata <- md$read_h5mu(h5mu_file)
```

Access individual modalities and convert them:

```{r access_modalities}
rna_mod <- mdata$mod[["rna"]]

rna_seurat <- rna_mod$as_Seurat()
print(rna_seurat)

rna_sce <- rna_mod$as_SingleCellExperiment()
print(rna_sce)
```

# Session info

## R

```{r r-sessioninfo}
sessionInfo()
```

## Python

```{r python-sessioninfo}
reticulate::py_config()

reticulate::py_list_packages()
```