### Install Tangram Package

Source: https://github.com/broadinstitute/tangram/blob/master/README.md

Instructions for installing the Tangram Python package using conda and pip. Requires PyTorch and scanpy to be pre-installed. Assumes the existence of an environment.yml file for conda environment setup.

```bash
conda env create -f environment.yml
conda activate tangram-env
pip install tangram-sc
```

--------------------------------

### Install Tangram via pip

Source: https://github.com/broadinstitute/tangram/blob/master/docs/source/getting_started.rst

Installs the tangram-sc package using pip. Ensure PyTorch and scanpy are installed beforehand. Refer to environment.yml for detailed dependencies.

```shell
pip install tangram-sc
```

--------------------------------

### Install and Import Tangram and Dependencies (Python)

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

This code block installs the necessary Python packages for the Tangram tutorial, including scanpy, torch, and tangram-sc. It also imports these libraries and sets up the plotting environment. Ensure 'tangram-sc' is installed via pip or by uncommenting the sys.path modification for local imports.

```python
import os, sys
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import scanpy as sc
import torch
sys.path.append('./')  # uncomment for local import
import tangram as tg

%load_ext autoreload
%autoreload 2
%matplotlib inline

tg.__version__
```

--------------------------------

### Setup Tangram Environment and Imports (Python)

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Sets up the Python environment for Tangram by importing necessary libraries such as scanpy, squidpy, numpy, pandas, anndata, and matplotlib. It also prints the versions of scanpy and squidpy to ensure compatibility. The %load_ext, %autoreload, and %matplotlib inline magic commands are used for interactive development.

```python
import scanpy as sc
import squidpy as sq
import numpy as np
import pandas as pd
from anndata import AnnData
import pathlib
import matplotlib.pyplot as plt
import matplotlib as mpl
import skimage
import seaborn as sns
import tangram as tg

sc.logging.print_header()
print(f"squidpy=={sq.__version__}")

%load_ext autoreload
%autoreload 2
%matplotlib inline
```

--------------------------------

### Tangram Mapping Training Output

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Displays the training progress and output messages from the Tangram mapping process. This includes information about tensor allocation, training start, and periodic score printing.

```log
INFO:root:Allocate tensors for mapping.
INFO:root:Begin training with 1280 genes and rna_count_based density_prior in cells mode...
INFO:root:Printing scores every 100 epochs.
```

--------------------------------

### Create and Activate Tangram Conda Environment

Source: https://github.com/broadinstitute/tangram/blob/master/docs/source/getting_started.rst

Creates a new conda environment from the provided environment.yml file and activates it. This ensures all necessary dependencies are correctly installed for Tangram.

```shell
conda env create --file environment.yml
conda activate tangram-env
```

--------------------------------

### Initialize Tangram Mapping

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Initializes the Tangram mapping process with specified parameters. It sets up the training configuration including cluster labels, density prior, number of epochs, and the device to be used for computation.

```python
import tangram as tg
import anndata
import scanpy as sc
import pandas as pd
import matplotlib.pyplot as plt

ad_map = tg.map_cells_to_space(
    adata_sc,
    adata_st,
    cluster_label='cell_subclass',  # .obs field w cell types
    density_prior='rna_count_based',
    num_epochs=500,
    # device="cuda:0",
    device='cpu',
)
```

--------------------------------

### Download and Unzip Spatial and Single-Cell Data (Shell)

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

This code snippet downloads the necessary datasets for the Tangram tutorial using wget. It fetches the single-cell RNA-seq data (mop_sn_tutorial.h5ad.gz), spatial data (slideseq_MOp_1217.h5ad.gz), and marker genes (MOp_markers.csv). It then unzips the .gz files. If wget is not available, manual download instructions are provided.

```shell
# Skip this cells if data are already downloaded
!wget https://storage.googleapis.com/tommaso-brain-data/tangram_demo/mop_sn_tutorial.h5ad.gz -O data/mop_sn_tutorial.h5ad.gz
!wget https://storage.googleapis.com/tommaso-brain-data/tangram_demo/slideseq_MOp_1217.h5ad.gz -O data/slideseq_MOp_1217.h5ad.gz
!wget https://storage.googleapis.com/tommaso-brain-data/tangram_demo/MOp_markers.csv -O data/MOp_markers.csv
!gunzip -f data/mop_sn_tutorial.h5ad.gz
!gunzip -f data/slideseq_MOp_1217.h5ad.gz
```

--------------------------------

### Cross-Validation for Hyperparameter Tuning

Source: https://context7.com/broadinstitute/tangram/llms.txt

Demonstrates the initial steps for setting up cross-validation in Tangram, such as loading and preprocessing single-cell (adata_sc) and spatial (adata_sp) data. This is a prerequisite for evaluating mapping quality and optimizing hyperparameters using techniques like leave-one-out or k-fold cross-validation.

```python
import scanpy as sc
import tangram as tg

# Load and preprocess data
adata_sc = sc.read_h5ad("single_cell_data.h5ad")
adata_sp = sc.read_h5ad("spatial_data.h5ad")
tg.pp_adatas(adata_sc, adata_sp, genes=None)
```

--------------------------------

### Import Tangram Library

Source: https://github.com/broadinstitute/tangram/blob/master/README.md

Demonstrates how to import the Tangram library into a Python script or Jupyter notebook for use in subsequent analyses.

```python
import tangram as tg
```

--------------------------------

### Load and Prepare Spatial and Single-Cell Data

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Loads spatial and single-cell data from H5AD files and preprocesses the single-cell data by normalizing total counts. It also reads gene marker data from a CSV file and reshapes it for Tangram.

```python
path = os.path.join('data', 'slideseq_MOp_1217.h5ad')
ad_sp = sc.read_h5ad(path)

path = os.path.join('data','mop_sn_tutorial.h5ad')
ad_sc = sc.read_h5ad(path)
sc.pp.normalize_total(ad_sc)

df_genes = pd.read_csv('data/MOp_markers.csv', index_col=0)
markers = np.reshape(df_genes.values, (-1, ))
markers = list(markers)

tg.pp_adatas(ad_sc, ad_sp, genes=markers)
```

--------------------------------

### AnnData Object Representation

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Shows the summarized information of the `AnnData` object returned by the mapping process. This includes the dimensions and a list of keys for `obs`, `var`, and `uns`.

```python
ad_map
```

--------------------------------

### Prepare AnnData Objects for Tangram Mapping

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Prepares single-cell (ad_sc) and spatial (ad_sp) AnnData objects for Tangram mapping. It subsets genes to be shared between both datasets, ensures gene order consistency, and converts gene names to lowercase by default. The number of training genes and overlapped genes are reported.

```python
tg.pp_adatas(ad_sc, ad_sp, genes=markers)
```

--------------------------------

### Import Libraries for Tangram Analysis

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Imports essential Python libraries including numpy, pandas, matplotlib, scanpy, torch, and tangram for data manipulation, plotting, and spatial transcriptomics analysis.

```python
import os, sys
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import scanpy as sc
import torch
import tangram as tg
```

--------------------------------

### Project Cell Annotations with Tangram

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Projects cell type annotations from a single-cell reference onto a spatial dataset using Tangram. It takes an annotation map, a spatial AnnData object, and the annotation key as input.

```python
tg.project_cell_annotations(ad_map, adata_st, annotation="cell_subclass")
```

--------------------------------

### AnnData Object Details

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Provides a detailed breakdown of the `AnnData` object's contents after mapping. It lists the number of observations and variables, along with the specific fields available in the `obs`, `var`, and `uns` attributes.

```text
Result:
AnnData object with n_obs × n_vars = 21697 × 324
    obs: 'sample_name', 'organism', 'donor_sex', 'cell_class', 'cell_subclass', 'cell_cluster', 'n_genes_by_counts', 'log1p_n_genes_by_counts', 'total_counts', 'log1p_total_counts', 'pct_counts_in_top_50_genes', 'pct_counts_in_top_100_genes', 'pct_counts_in_top_200_genes', 'pct_counts_in_top_500_genes', 'total_counts_mt', 'log1p_total_counts_mt', 'pct_counts_mt', 'n_counts'
    var: 'in_tissue', 'array_row', 'array_col', 'n_genes_by_counts', 'log1p_n_genes_by_counts', 'total_counts', 'log1p_total_counts', 'pct_counts_in_top_50_genes', 'pct_counts_in_top_100_genes', 'pct_counts_in_top_200_genes', 'pct_counts_in_top_500_genes', 'total_counts_MT', 'log1p_total_counts_MT', 'pct_counts_MT', 'n_counts', 'leiden', 'cluster', 'uniform_density', 'rna_count_based_density'
    uns: 'train_genes_df', 'training_history'
```

--------------------------------

### Visualize Cell Count with Scanpy

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Adds cell count data to the AnnData object's observation DataFrame and visualizes spatial data using Scanpy's plotting function. It allows coloring by cluster and cell count.

```python
adata_st.obs["cell_count"] = adata_st.obsm["image_features"]["segmentation_label"]
sc.pl.spatial(adata_st, color=["cluster", "cell_count"], frameon=False)
```

--------------------------------

### Explanation of Training Scores Plot Panels

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Provides a detailed explanation of the four panels within the `plot_training_scores` output. It describes what each panel visualizes, including histograms of similarity scores, gene sparsity comparisons between single-cell and spatial data, and the relationship between training scores and sparsity differences.

```text
- The first panel is a histogram of the simlarity scores for each training gene.
- In the second panel, each dot is a training gene and we can observe the training score (y-axis) and the sparsity in the scRNA-seq data (x-axis) of each gene. 
- The third panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data are usually more sparse than single cell data, a discrepancy which is often responsible for low quality mapping.
- The last panel is similar to the second one, but contains the gene sparsity of the spatial data. Spatial data
```

--------------------------------

### Visualizing Segmentation Results with Matplotlib and Squidpy

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Visualizes cell segmentation masks alongside the original DAPI staining and cluster information. This helps in assessing the quality of the segmentation by comparing it with the histology.

```python
inset_y = 1500
inset_x = 1700
inset_sy = 400
inset_sx = 500

fig, axs = plt.subplots(1, 3, figsize=(30, 10))
sc.pl.spatial(
    adata_st, color="cluster", alpha=0.7, frameon=False, show=False, ax=axs[0], title=""
)
axs[0].set_title("Clusters", fontdict={"fontsize": 20})
sf = adata_st.uns["spatial"]["V1_Adult_Mouse_Brain_Coronal_Section_2"]["scalefactors"][
    "tissue_hires_scalef"
]
rect = mpl.patches.Rectangle(
    (inset_y * sf, inset_x * sf),
    width=inset_sx * sf,
    height=inset_sy * sf,
    ec="yellow",
    lw=4,
    fill=False,
)
axs[0].add_patch(rect)

axs[0].axes.xaxis.label.set_visible(False)
axs[0].axes.yaxis.label.set_visible(False)

axs[1].imshow(
    img["image"][inset_y : inset_y + inset_sy, inset_x : inset_x + inset_sx, 0, 0] / 65536,
    interpolation="none",
)
axs[1].grid(False)
axs[1].set_xticks([])
axs[1].set_yticks([])
axs[1].set_title("DAPI", fontdict={"fontsize": 20})

crop = img["segmented_watershed"](
    inset_y : inset_y + inset_sy, inset_x : inset_x + inset_sx
).values.squeeze(-1)
crop = skimage.segmentation.relabel_sequential(crop)[0]
cmap = plt.cm.plasma
cmap.set_under(color="black")
axs[2].imshow(crop, interpolation="none", cmap=cmap, vmin=0.001)
axs[2].grid(False)
axs[2].set_xticks([])
axs[2].set_yticks([])
axs[2].set_title("Nucleous segmentation", fontdict={"fontsize": 20});
```

--------------------------------

### Load Spatial and Single-Cell Datasets with Squidpy

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Loads spatial and single-cell datasets from Squidpy. It includes subsetting spatial data to specific cortex clusters and cropping the associated image. The single-cell data is sourced from Tasic et al. (2018) and pre-processed using scanpy.

```python
adata_st = sq.datasets.visium_fluo_adata_crop()
adata_st = adata_st[
    adata_st.obs.cluster.isin([f"Cortex_{i}" for i in np.arange(1, 5)])
].copy()
img = sq.datasets.visium_fluo_image_crop()

adata_sc = sq.datasets.sc_mouse_cortex()
```

--------------------------------

### Image Pre-processing and Segmentation with Squidpy

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Applies smoothing to an image as a pre-processing step and then computes segmentation masks using the watershed algorithm. This is crucial for identifying individual cells within spatial voxels.

```python
sq.im.process(img=img, layer="image", method="smooth")
sq.im.segment(
    img=img,
    layer="image_smooth",
    method="watershed",
    channel=0,
)
```

--------------------------------

### Project Cell Annotations and Plot

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Projects cell annotations from the single-cell data onto the spatial data using Tangram and then plots the cell annotations in space. This involves transferring the 'cell_subclass' annotation and visualizing it using `plot_cell_annotation_sc`.

```python
tg.project_cell_annotations(ad_map, adata_st, annotation="cell_subclass")
annotation_list = list(pd.unique(adata_sc.obs['cell_subclass']))
tg.plot_cell_annotation_sc(adata_st, annotation_list,perc=0.02)
```

--------------------------------

### Load Marker Genes for Tangram Mapping

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Loads marker genes from a CSV file into a pandas DataFrame and extracts them as a list. This list will be used as training genes for Tangram.

```python
df_genes = pd.read_csv('data/MOp_markers.csv', index_col=0)
markers = np.reshape(df_genes.values, (-1, ))
markers = list(markers)
len(markers)
```

--------------------------------

### Display Spatial Dataset Metadata

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Displays the observation (metadata) of the spatial dataset (`adata_st`). This output shows various metrics associated with each spatial coordinate, such as tissue status, gene counts, and mitochondrial content.

```python
adata_st.obs
```

--------------------------------

### Select and Display Gene Data

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Selects specific genes from the `df_all_genes` DataFrame and displays their associated data, including score, training status, and sparsity metrics. This is useful for inspecting individual gene performance.

```python
genes = ['snap25', 'atp1b1', 'atp1a3', 'ctgf', 'nefh', 'aak1', 'fa2h', ]
df_all_genes.loc[genes]
```

--------------------------------

### Create Segmentation Cell Dataframe with Tangram

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Creates a dataframe mapping segmentation objects (cells) to spots in the spatial data. It extracts spatial coordinates and generates unique centroid IDs, storing the result in the AnnData object's 'uns' attribute.

```python
tg.create_segment_cell_df(adata_st)
adata_st.uns["tangram_cell_segmentation"].head()
```

--------------------------------

### Mapper.train Method

Source: https://github.com/broadinstitute/tangram/blob/master/docs/source/classes/tangram.mapping_optimizer.Mapper.train.rst

This method is used for training a mapper object. It likely takes data and configuration as input to optimize the mapping process.

```APIDOC
## Mapper.train

### Description
Trains the mapper object using the provided data and configuration.

### Method
*This is a method call within a class, not a direct HTTP endpoint.*

### Endpoint
*N/A - Class method*

### Parameters
*Detailed parameters for this method would typically be found in the Python docstring or source code.*

### Request Example
*N/A - Python method call*

### Response
*N/A - Python method return value would be the trained mapper object.*

#### Success Response (200)
*N/A*

#### Response Example
*N/A*
```

--------------------------------

### Tangram Training Scores Output

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Shows the iterative training scores and KL regularization values during the Tangram mapping process. These scores are printed every 100 epochs, providing insight into the training convergence.

```log
Score: 0.613, KL reg: 0.001
Score: 0.733, KL reg: 0.000
Score: 0.736, KL reg: 0.000
Score: 0.737, KL reg: 0.000
Score: 0.737, KL reg: 0.000
```

--------------------------------

### Load Single Cell Data with Scanpy

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Loads single-cell RNA sequencing (scRNAseq) or single-nuclei RNA sequencing (snRNAseq) data from an h5ad file into an AnnData object. This snippet requires the scanpy library.

```python
import scanpy as sc
import os

path = os.path.join('data','mop_sn_tutorial.h5ad')
ad_sc = sc.read_h5ad(path)
ad_sc
```

--------------------------------

### Constrained Deconvolution with Tangram

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Performs cell deconvolution to map single-cell profiles to spatial locations using Tangram. This method uses a 'constrained' mode to match the number of segmented cells and requires a density prior.

```python
ad_map = tg.map_cells_to_space(
    adata_sc,
    adata_st,
    mode="constrained",
    target_count=adata_st.obs.cell_count.sum(),
    density_prior=np.array(adata_st.obs.cell_count) / adata_st.obs.cell_count.sum(),
    num_epochs=1000,
)
```

--------------------------------

### Visualize Gene Patterns with Tangram

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Uses the `plot_genes_sc` function from the Tangram library to visualize gene expression patterns. It compares spatial measurements (`ad_sp`) with predicted gene expression (`ad_ge`) for a given list of genes. Parameters like `spot_size`, `scale_factor`, and `perc` control the visualization's appearance and data filtering.

```python
tg.plot_genes_sc(genes, adata_measured=ad_sp, adata_predicted=ad_ge, spot_size=50, scale_factor=0.1, perc=0.001, return_figure=False)
```

--------------------------------

### Project Cell Annotations onto Spatial Data

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

This function projects cell type annotations from single-cell data onto the spatial data using the Tangram mapping results. It helps visualize the spatial distribution of cell types. Dependencies include pandas for unique value extraction.

```python
tg.project_cell_annotations(ad_map, ad_sp, annotation='subclass_label')
annotation_list = list(pd.unique(ad_sc.obs['subclass_label']))
tg.plot_cell_annotation_sc(ad_sp, annotation_list,x='x', y='y',spot_size= 60, scale_factor=0.1, perc=0.001)
```

--------------------------------

### AnnData Structure of Mapping Results

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Describes the structure of the AnnData object (`ad_map`) that stores the mapping results. It details the contents of the `X` matrix, `obs` dataframe, `var` dataframe, and `uns` dictionary, explaining what each component represents.

```text
The mapping results are stored in the returned `AnnData` structure, saved as `ad_map`, structured as following:
- The cell-by-spot matrix `X` contains the probability of cell `i` to be in spot `j`.
- The `obs` dataframe contains the metadata of the single cells.
- The `var` dataframe contains the metadata of the spatial data.
- The `uns` dictionary contains a dataframe with various information about the training genes (saved as `train_genes_df`).
```

--------------------------------

### Load Spatial Data with Scanpy

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_without_squidpy.ipynb

Loads spatial data from an h5ad file into an AnnData object. The data is expected to be in a voxel-by-gene matrix format. This snippet requires the scanpy library.

```python
import scanpy as sc
import os

path = os.path.join('./data', 'slideseq_MOp_1217.h5ad')
ad_sp = sc.read_h5ad(path)
ad_sp
```

--------------------------------

### Feature Extraction for Deconvolution with Tangram

Source: https://github.com/broadinstitute/tangram/blob/master/tutorial_tangram_with_squidpy.ipynb

Defines keyword arguments for extracting image features essential for deconvolution. It specifies the segmentation layer to use and the properties to extract, such as 'label' and 'centroid'.

```python
# define image layer to use for segmentation
features_kwargs = {
    "segmentation": {
        "label_layer": "segmented_watershed",
        "props": ["label", "centroid"],
        "channels": [1, 2],
    }
}
```