### Install BANKSY from Source for Development

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Clone the BANKSY repository and install it in editable mode for development purposes.

```bash
git clone https://github.com/prabhakarlab/Banksy_py.git
cd Banksy_py
pip install -e .
```

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

### Install BANKSY from GitHub

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Install the BANKSY package directly from its GitHub repository.

```bash
pip install git+https://github.com/prabhakarlab/Banksy_py.git
```

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

### Install Optional Dependencies from Source for BANKSY

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Install optional dependencies for BANKSY when installing from source. Use '.[notebooks]', '.[mclust]', or '.[all]'.

```bash
pip install -e ".[notebooks]"
```

```bash
pip install -e ".[mclust]"
```

```bash
pip install -e ".[all]"
```

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

### Install All Optional Dependencies for BANKSY

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Use this command to install BANKSY with all optional dependencies, including mclust and Jupyter.

```bash
pip install "pybanksy[all]"
```

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

### Install Harmony Integration for BANKSY

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Install the Harmony PyTorch package using conda, required for running the harmony integration example notebook.

```bash
conda install bioconda::harmony-pytorch
```

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

### Import Libraries and Setup

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv1_analysis.ipynb

Imports necessary libraries and configures environment settings for data analysis. Sets random seeds for reproducibility.

```python
import os, re
import numpy as np
import pandas as pd
from IPython.display import display
import warnings
warnings.filterwarnings("ignore") 

import scipy.sparse as sparse
from scipy.io import mmread
from scipy.stats import pearsonr, pointbiserialr

import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
from matplotlib.lines import Line2D

import seaborn as sns
import scanpy as sc
sc.logging.print_header()
sc.set_figure_params(facecolor="white", figsize=(8, 8))
sc.settings.verbosity = 1 # errors (0), warnings (1), info (2), hints (3)
plt.rcParams["font.family"] = "Arial"
sns.set_style("white")

import random
# Note that BANKSY itself is deterministic, here the seeds affect the umap clusters and leiden partition
seed = 1234
np.random.seed(seed)
random.seed(seed)
```

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

### Install pybanksy

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Install the pybanksy package using pip. Optional dependencies can be included for notebooks, mclust, or all functionalities.

```bash
pip install pybanksy
```

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

### BANKSY Analysis Setup and Execution

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

This snippet outlines the initial setup and execution of the BANKSY analysis. It shows the number of genes analyzed, gene list, and the creation of the BANKSY matrix. It also details the decay types and dimensionality reduction steps.

```python
results_df_modi
```

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

### Import Libraries and Setup Environment

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv2_analysis.ipynb

Imports essential Python libraries for data manipulation, scientific computing, plotting, and spatial transcriptomics analysis. Sets up Scanpy's logging and figure parameters, and initializes random seeds for reproducibility.

```python
import os, re
import numpy as np
import pandas as pd
from IPython.display import display
import warnings
warnings.filterwarnings("ignore") 

import scipy.sparse as sparse
from scipy.io import mmread
from scipy.stats import pearsonr, pointbiserialr

import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
from matplotlib.lines import Line2D

import seaborn as sns
import scanpy as sc
sc.logging.print_header()
sc.set_figure_params(facecolor="white", figsize=(8, 8))
sc.settings.verbosity = 1 # errors (0), warnings (1), info (2), hints (3) 
plt.rcParams["font.family"] = "Arial"
sns.set_style("white")

import random
# Note that BANKSY itself is deterministic, here the seeds affect the umap clusters and leiden partition
seed = 1234
np.random.seed(seed)
random.seed(seed)
```

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

### Install Jupyter Notebook Optional Dependency for BANKSY

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Use this command to install BANKSY with Jupyter notebook support.

```bash
pip install "pybanksy[notebooks]"
```

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

### Quick Start: Initialize and Run BANKSY Clustering

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Load spatial transcriptomics data, initialize BANKSY with specified parameters, and run clustering with multiple lambda and resolution values.

```python
import scanpy as sc
from banksy.initialize_banksy import initialize_banksy
from banksy.run_banksy import run_banksy_multiparam

# Load your spatial transcriptomics data
adata = sc.read_h5ad("your_data.h5ad")

# Initialize BANKSY
coord_keys = ('x', 'y', 'spatial')  # Adjust based on your data
banksy_dict = initialize_banksy(
    adata,
    coord_keys=coord_keys,
    num_neighbours=15,
    nbr_weight_decay='scaled_gaussian'
)

# Run BANKSY clustering

results_df = run_banksy_multiparam(
    adata,
    banksy_dict,
    lambda_list=[0.2],
    resolutions=[0.5, 1.0]
)
```

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

### Install mclust Optional Dependency for BANKSY

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Use this command to install BANKSY with mclust clustering support. Requires R and rpy2.

```bash
pip install "pybanksy[mclust]"
```

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

### Run Leiden Partition for Cell Clustering

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv1_analysis.ipynb

Clusters cells using the Leiden algorithm. Recommended for cases where the number of clusters is unknown. Ensure 'leidenalg' is installed if needed.

```python
from banksy.cluster_methods import run_Leiden_partition

results_df, max_num_labels = run_Leiden_partition(
    banksy_dict,
    resolutions,
    num_nn = 50,
    num_iterations = -1,
    partition_seed = seed,
    match_labels = True,
)
```

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

### Scanpy Version Information

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B006_ascending.ipynb

Displays the installed versions of key libraries used in the analysis, including scanpy, anndata, and others.

```text
scanpy==1.9.5 anndata==0.10.2 umap==0.5.4 numpy==1.24.4 scipy==1.11.3 pandas==2.1.1 scikit-learn==1.3.1 statsmodels==0.14.0 igraph==0.10.6 pynndescent==0.5.10
```

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

### Import Libraries

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_concatenate_multisample.ipynb

Imports common libraries used for data analysis, including pandas, numpy, scanpy, and scipy. Ensure these libraries are installed.

```python
import os,csv,re
import pandas as pd
import numpy as np

import math
from scipy.sparse import issparse
import random 
import warnings
warnings.filterwarnings("ignore")

import matplotlib.colors as clr
import matplotlib.pyplot as plt

import scanpy as sc

# opencv for reading in images only
# import cv2

import time
```

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

### PCA and UMAP with Scaled Gaussian Decay

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_concatenate_multisample.ipynb

This snippet demonstrates the setup and execution of PCA and UMAP with scaled Gaussian decay. It includes nearest-neighbor graph construction, shared nearest-neighbor analysis, and graph partitioning. The output shows the modularity and unique labels found.

```python
Output:
Decay type: scaled_gaussian
Neighbourhood Contribution (Lambda Parameter): 0.2
reduced_pc_20 

reduced_pc_20_umap 

PCA dims to analyse: [20]

====================================================================================================
Setting up partitioner for (nbr decay = scaled_gaussian), Neighbourhood contribution = 0.2, PCA dimensions = 20)
====================================================================================================


Nearest-neighbour weighted graph (dtype: float64, shape: (14243, 14243)) has 712150 nonzero entries.
---- Ran find_nn in 1.83 s ----


Nearest-neighbour connectivity graph (dtype: int16, shape: (14243, 14243)) has 712150 nonzero entries.

(after computing shared NN)
Allowing nearest neighbours only reduced the number of shared NN from 12633261 to 710946.


Shared nearest-neighbour (connections only) graph (dtype: int16, shape: (14243, 14243)) has 683775 nonzero entries.

Shared nearest-neighbour (number of shared neighbours as weights) graph (dtype: int16, shape: (14243, 14243)) has 683775 nonzero entries.

sNN graph data:
[24 19 10 ...  7  6  6]

---- Ran shared_nn in 0.45 s ----


-- Multiplying sNN connectivity by weights --


shared NN with distance-based weights graph (dtype: float64, shape: (14243, 14243)) has 683775 nonzero entries.
shared NN weighted graph data: [0.19646475 0.19658107 0.19720675 ... 0.12681301 0.13242565 0.1369962 ]

Converting graph (dtype: float64, shape: (14243, 14243)) has 683775 nonzero entries.
---- Ran csr_to_igraph in 0.10 s ----


Resolution: 0.6
------------------------------

---- Partitioned BANKSY graph ----
modularity: 0.75
7 unique labels:
[0 1 2 3 4 5 6]

---- Ran partition in 3.40 s ----

No annotated labels
```

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

### Load Manual Annotations

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

Loads and integrates manual cell type annotations into the AnnData object. This step is crucial for validating or guiding the clustering results.

```python
'''Load manual annotations'''
adata = adata[adata.obs["cluster_name"].notnull()]

annotations =  pd.read_csv(os.path.join(file_path, "Starmap_BY3_1k_meta_annotated_18oct22.csv"))

manual_labels =  "smoothed_manual" # Ground truth annotations in pd.DataFrame
annotation_key = 'manual_annotations' # Key to access annotations in adata.obs[annotation_keys]

print(annotations.loc[:,manual_labels ])
adata.obs[annotation_key] = annotations.loc[:,manual_labels ].values
adata.obs[annotation_key] = adata.obs[annotation_key].astype('category')
print(adata.obs[annotation_key])

# Add spatial coordinates to '.obsm' attribute
adata.obsm[coord_keys[2]] = pd.concat([adata.obs[coord_keys[0]], adata.obs[coord_keys[1]]], axis=1).to_numpy()
```

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

### Initialize BANKSY and Scanpy Environment

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B012_tissue_segmentation.ipynb

Sets up the Python environment by importing necessary libraries and configuring scanpy settings for data analysis. Includes random seed for reproducibility.

```python
# This is the simplified python script that demonstrates how to directly run slide-seq v1 dataset with BANKSY
import os, time, random, gc
import anndata
from anndata import AnnData
import numpy as np
import pandas as pd
import warnings
from banksy_utils.color_lists import spagcn_color
warnings.filterwarnings("ignore")
import scanpy as sc
sc.logging.print_header()
sc.set_figure_params(facecolor="white", figsize=(8, 8))
sc.settings.verbosity = 1  # errors (0), warnings (1), info (2), hints (3)

seed = 0
np.random.seed(seed)
random.seed(seed)
start = time.perf_counter_ns()
```

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

### Initialize BANKSY Parameters

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_harmony_multisample.ipynb

Sets up initial parameters for BANKSY, including coordinate keys, neighbor weight decay, and geometric parameters like k_geom.

```python
## banksy parameters ##
from banksy.initialize_banksy import initialize_banksy
coord_keys = ('x_pixel', 'y_pixel', 'coord_xy')
nbr_weight_decay = 'scaled_gaussian'
k_geom = 18
x_coord, y_coord, xy_coord = coord_keys[0], coord_keys[1], coord_keys[2]
```

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

### Get Python Version

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_concatenate_multisample.ipynb

Retrieves the current Python version. Useful for environment checks.

```python
import platform
platform.python_version()
```

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

### Label Expansion and Matching

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_concatenate_multisample.ipynb

This snippet details the post-partitioning steps of expanding labels to cover a full range and then matching them. It shows the resulting label IDs after these operations.

```python
After sorting Dataframe
Shape of dataframe: (2, 7)

Maximum number of labels = 7
Indices of sorted list: [0 1]

Expanding labels with ids: [0 1 2 3 4 5 6] so that ids range from 0 to 6

Label ids zerod: [0 1 2 3 4 5 6].
0 to be inserted between each id: [0 0 0 0 0 0 0]
0 extra rows to be randomly inserted: [0 0 0 0 0 0]
New ids: [0 1 2 3 4 5 6]
---- Ran expand_labels in 0.00 s ----

---- Ran match_labels in 0.00 s ----


Matched Labels
```

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

### Initialize Banksy and Set Random Seeds

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

Imports necessary libraries and sets up random seeds for reproducibility. This code should be run at the beginning of the analysis.

```python
import warnings
warnings.filterwarnings("ignore") 
import os, time
import pandas as pd
import numpy as np
import random
import scipy.sparse as sparse
from scipy.sparse import csr_matrix, issparse

from banksy.initialize_banksy import initialize_banksy
from banksy.run_banksy import run_banksy_multiparam
from banksy_utils.color_lists import spagcn_color

start = time.perf_counter_ns() 
random_seed = 1234
cluster_algorithm = 'leiden'
np.random.seed(random_seed)
random.seed(random_seed)
```

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

### Display AnnData Object Summary

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv2_analysis.ipynb

Use the `display_adata` function to get a summary of an AnnData object, including its dimensions, sparsity, and attributes for observations and variables.

```python
display_adata(adata)
```

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

### BANKSY Initialization Parameters

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B012_tissue_segmentation.ipynb

Sets up default parameters for BANKSY, including geometric parameters, weighting decay, clustering resolutions, PCA dimensions, and lambda values.

```python
from banksy.main import median_dist_to_nearest_neighbour
from banksy.initialize_banksy import initialize_banksy
from banksy.embed_banksy import generate_banksy_matrix
from banksy.main import concatenate_all

k_geom = 15  # only for fixed type
max_m = 1  # azumithal transform up to kth order
nbr_weight_decay = "scaled_gaussian"  # can also be "reciprocal", "uniform" or "ranked"
resolutions = list(np.arange(0.08, 0.1, 0.005))#[0.08]  # clustering resolution for leiden algorithm
max_labels = None # Number of clusters for tissue segmentation
pca_dims = [20]  # Dimensionality in which PCA reduces to
lambda_list = [0.8]
```

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

### Initializing BANKSY Analysis Parameters

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

Initialize the BANKSY dictionary with various parameters for spatial graph construction and analysis. This includes setting the number of neighbors, decay method, and plotting options.

```python
resolutions = [.9] # clustering resolution for Leiden clustering

pca_dims = [20] # number of dimensions to keep after PCA

lambda_list = [.8] # lambda

k_geom = 8 #spatial neighbours

max_m = 1 # use AGF

nbr_weight_decay = "scaled_gaussian" # can also be "reciprocal", "uniform" or "ranked"

```

```python
banksy_dict_modi = initialize_banksy(
    adata,
    coord_keys,
    k_geom,
    nbr_weight_decay=nbr_weight_decay,
    max_m=max_m,
    plt_edge_hist=True,
    plt_nbr_weights=True,
    plt_agf_angles=False,
    plt_theta=False,
)
```

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

### Import BANKSY Libraries and Set Parameters

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B006_ascending.ipynb

Imports necessary functions from the BANKSY library and defines default parameters for domain segmentation. These parameters include k_geom, max_m, nbr_weight_decay, resolutions, max_labels, pca_dims, and lambda_list.

```python
from banksy.main import median_dist_to_nearest_neighbour
from banksy.initialize_banksy import initialize_banksy
from banksy.embed_banksy import generate_banksy_matrix
from banksy.main import concatenate_all

k_geom = 15  # only for fixed type
max_m = 1  # azumithal transform up to kth order
br_weight_decay = "scaled_gaussian"  # can also be "reciprocal", "uniform" or "ranked"
resolutions = None  # clustering resolution for leiden algorithm
max_labels = 8 # Number of clusters for tissue segmentation
pca_dims = [20]  # Dimensionality in which PCA reduces to
lambda_list = [0.8]
```

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

### Generate BANKSY Embedding Matrix

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_harmony_multisample.ipynb

Generates the BANKSY embedding matrix using specified lambda parameters and maximum order 'm'. Includes setup for PCA dimensions and clustering resolutions.

```python
from banksy.main import concatenate_all
from banksy.embed_banksy import generate_banksy_matrix
resolutions = [0.40] # clustering resolution for UMAP
pca_dims = [20] # Dimensionality in which PCA reduces to
lambda_list = [0.2] # list of lambda parameters
m = 1
from banksy_utils.umap_pca import pca_umap
from banksy.cluster_methods import run_Leiden_partition
from banksy.plot_banksy import plot_results
c_map =  'tab20' # specify color map


### run banksy ###


# Include spatial coordinates information
raw_y, raw_x = adata.obs[y_coord], adata.obs[x_coord]
adata.obsm[xy_coord] = np.vstack((adata.obs[x_coord].values, adata.obs[y_coord].values)).T

banksy_dict = initialize_banksy(adata,
                                coord_keys,
                                k_geom,
                                nbr_weight_decay=nbr_weight_decay,
                                max_m=m,
                                plt_edge_hist= True,
                                plt_nbr_weights= True,
                                plt_agf_angles=False,
                                plt_theta=False
                               )

banksy_dict, banksy_matrix = generate_banksy_matrix(adata,
                                                    banksy_dict,
                                                    lambda_list,
                                                    max_m=m)
```

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

### BANKSY Analysis Results DataFrame

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

This is an example of the results dataframe generated by BANKSY after an initial analysis. It includes parameters used, clustering metrics like ARI, and references to the data object.

```text
Result:
                                             decay  lambda_param  num_pcs  
scaled_gaussian_pc20_nc0.80_r0.90  scaled_gaussian           0.8       20   

                                   resolution  num_labels  
scaled_gaussian_pc20_nc0.80_r0.90         0.9           7   

                                                                              labels  
scaled_gaussian_pc20_nc0.80_r0.90  Label object:\nNumber of labels: 7, number of ...   

                                                                               adata  
scaled_gaussian_pc20_nc0.80_r0.90  [[[View of AnnData object with n_obs × n_vars ...   

                                        ari  
scaled_gaussian_pc20_nc0.80_r0.90  0.755481  
```

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

### Identify Mitochondrial Genes in AnnData

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv1_analysis.ipynb

Adds a boolean column 'mt' to the AnnData object's variable attributes, indicating whether a gene name starts with 'MT-'. This is used for filtering or QC.

```python
adata.var["mt"] = adata.var_names.str.startswith("MT-")
```

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

### Initialize Banksy with Visualization Options

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv2_analysis.ipynb

Initializes the banksy object with specified parameters and enables several visualization options for edge histograms, neighbor weights, and theta plots. Use this to explore the spatial characteristics of your data.

```python
from banksy.initialize_banksy import initialize_banksy

banksy_dict = initialize_banksy(
    adata,
    coord_keys,
    k_geom,
    nbr_weight_decay=nbr_weight_decay,
    max_m=max_m,
    plt_edge_hist=True,
    plt_nbr_weights=True,
    plt_agf_angles=False, # takes long time to plot
    plt_theta=True,
)
```

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

### Run Leiden Partition and Plot Results

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B006_ascending.ipynb

This snippet demonstrates how to perform Leiden clustering and visualize the results. It requires pre-defined data structures like `banksy_dict`, `resolutions`, `coord_keys`, and `file_path`. Ensure all necessary imports are available.

```python
from banksy.cluster_methods import run_Leiden_partition

banksy_df, max_num_labels = run_Leiden_partition(
    banksy_dict,
    resolutions,
    num_nn = 50,
    num_iterations = -1,
    partition_seed = 1234,
    match_labels = True,
    max_labels = max_labels,
)

from banksy.plot_banksy import plot_results

c_map =  'tab20' # specify color map
weights_graph =  banksy_dict['scaled_gaussian']['weights'][1]

plot_results(
    banksy_df,
    weights_graph,
    c_map,
    match_labels = True,
    coord_keys = coord_keys,
    max_num_labels  =  max_num_labels, 
    save_path = os.path.join(file_path, 'BANKSY-Results'),
    save_fig = False, # Save Spatial Plot Only
    save_fullfig = True, # Save Full Plot
    dataset_name = f"CODEX-{unique_regions}",
    save_labels=True
)
banksy_df.to_csv(os.path.join(file_path, f"CODEX-{unique_regions}_BANKSY.csv"))
```

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

### Label Expansion and Sorting

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B006_ascending.ipynb

This section details the process of expanding and sorting labels after partitioning. It identifies the resolution and handles cases with no annotated labels, preparing the data for further analysis.

```python
Resolution found: 0.2 for labels: 8

Resolution: 0.2
------------------------------

---- Partitioned BANKSY graph ----
modularity: 0.75
7 unique labels:
[0 1 2 3 4 5 6]

---- Ran partition in 4.32 s ----

No annotated labels

After sorting Dataframe
Shape of dataframe: (1, 7)

Maximum number of labels = 7
Indices of sorted list: [0]

Expanding labels with ids: [0 1 2 3 4 5 6] so that ids range from 0 to 6

Label ids zerod: [0 1 2 3 4 5 6].
0 to be inserted between each id: [0 0 0 0 0 0 0]
0 extra rows to be randomly inserted: [0 0 0 0 0 0]
New ids: [0 1 2 3 4 5 6]
---- Ran expand_labels in 0.00 s ----
```

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

### Initialize Weighted Neighborhood Graphs

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B012_tissue_segmentation.ipynb

Initializes the weighted neighborhood graphs for BANKSY using median distance to nearest neighbors and various plotting options.

```python
nbrs = median_dist_to_nearest_neighbour(adata, key=coord_keys[2])
banksy_dict = initialize_banksy(adata,
                                coord_keys,
                                k_geom,
                                nbr_weight_decay=nbr_weight_decay,
                                max_m=max_m,
                                plt_edge_hist=False,
                                plt_nbr_weights=True,
                                plt_agf_angles=False,
                                plt_theta=False
                                )
```

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

### Initialize BANKSY

Source: https://github.com/prabhakarlab/banksy_py/blob/main/README.md

Initializes BANKSY to generate the spatial graph. This is a core step in the BANKSY pipeline.

```python
banksy.initialise_banksy(adata, banksy_dict)
```

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

### Preview Preprocessed Data

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv1_analysis.ipynb

Display the first few rows and columns of the preprocessed data to understand its structure and content. This is useful for a quick data sanity check after preprocessing.

```python
Result:
                    xcoord       ycoord  n_genes_by_counts  \
barcodes                                                     
TACTAAAGAATTA  2224.000000  1230.397849                 35   
TCTCTTAGTTGGC  1914.742424  2221.954545                 33   
GCTTTTCGTTCCC  1682.151515  3886.575758                 57   
TACGGGCGAAAAG  2416.261194  4216.902985                 14   
GTACCCTTCCGGG  3913.880342  1899.717949                283   
...                    ...          ...                ...   
CTACCGTGCGGCG  5176.223776  3567.398601                 30   
CCGATATGCGGCG  2222.324324  3906.189189                 48   
TTGGTATCGCCGC  3187.536082  2031.268041                 55   
TTCTTATCGCCGC  3686.817460  3833.047619                 75   
TTGTCGTATAGCG  4084.235294  1917.862745                106   

               log1p_n_genes_by_counts  total_counts  log1p_total_counts  \
barcodes                                                                   
TACTAAAGAATTA                 3.583519          39.0            3.688879   
TCTCTTAGTTGGC                 3.526361          38.0            3.663562   
GCTTTTCGTTCCC                 4.060443          63.0            4.158883   
TACGGGCGAAAAG                 2.708050          16.0            2.833213   
GTACCCTTCCGGG                 5.648974         345.0            5.846439   
...                                ...           ...                 ...   
CTACCGTGCGGCG                 3.433987          33.0            3.526361   
CCGATATGCGGCG                 3.891820          49.0            3.912023   
TTGGTATCGCCGC                 4.025352          58.0            4.077538   
TTCTTATCGCCGC                 4.330733          85.0            4.454347   
TTGTCGTATAGCG                 4.672829         113.0            4.736198   

               pct_counts_in_top_50_genes  pct_counts_in_top_100_genes  \
barcodes                                                                 
TACTAAAGAATTA                  100.000000                   100.000000   
TCTCTTAGTTGGC                  100.000000                   100.000000   
GCTTTTCGTTCCC                   88.888889                   100.000000   
TACGGGCGAAAAG                  100.000000                   100.000000   
GTACCCTTCCGGG                   32.463768                    46.956522   
...                                   ...                          ...   
CTACCGTGCGGCG                  100.000000                   100.000000   
CCGATATGCGGCG                  100.000000                   100.000000   
TTGGTATCGCCGC                   91.379310                   100.000000   
TTCTTATCGCCGC                   70.588235                   100.000000   
TTGTCGTATAGCG                   50.442478                    94.690265   

               pct_counts_in_top_200_genes  pct_counts_in_top_500_genes  \
barcodes                                                                  
TACTAAAGAATTA                   100.000000                        100.0   
TCTCTTAGTTGGC                   100.000000                        100.0   
GCTTTTCGTTCCC                   100.000000                        100.0   
TACGGGCGAAAAG                   100.000000                        100.0   
GTACCCTTCCGGG                    75.942029                        100.0   
...                                    ...                          ...   
CTACCGTGCGGCG                   100.000000                        100.0   
CCGATATGCGGCG                   100.000000                        100.0   
TTGGTATCGCCGC                   100.000000                        100.0   
TTCTTATCGCCGC                   100.000000                        100.0   
TTGTCGTATAGCG                   100.000000                        100.0   

               total_counts_mt  log1p_total_counts_mt  pct_counts_mt  
barcodes                                                              
TACTAAAGAATTA              0.0                    0.0            0.0  
TCTCTTAGTTGGC              0.0                    0.0            0.0  
GCTTTTCGTTCCC              0.0                    0.0            0.0  
TACGGGCGAAAAG              0.0                    0.0            0.0  
GTACCCTTCCGGG              0.0                    0.0            0.0  
...                        ...                    ...            ...  
CTACCGTGCGGCG              0.0                    0.0            0.0  
CCGATATGCGGCG              0.0                    0.0            0.0  
TTGGTATCGCCGC              0.0                    0.0            0.0  
TTCTTATCGCCGC              0.0                    0.0            0.0  
TTGTCGTATAGCG              0.0                    0.0            0.0  

[25551 rows x 13 columns]
```

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

### Harmony Partitioning with Scaled Gaussian Decay

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_harmony_multisample.ipynb

Sets up and runs graph partitioning using a scaled Gaussian decay type and a neighborhood contribution of 0.2 with 20 PCA dimensions. This snippet demonstrates the process of building weighted and connectivity graphs and performing community detection.

```python
Output:
Decay type: scaled_gaussian
Neighbourhood Contribution (Lambda Parameter): 0.2
reduced_pc_20 

reduced_pc_20_umap 

PCA dims to analyse: [20]

====================================================================================================
Setting up partitioner for (nbr decay = scaled_gaussian), Neighbourhood contribution = 0.2, PCA dimensions = 20)
====================================================================================================


Nearest-neighbour weighted graph (dtype: float64, shape: (11468, 11468)) has 573400 nonzero entries.
---- Ran find_nn in 1.49 s ----


Nearest-neighbour connectivity graph (dtype: int16, shape: (11468, 11468)) has 573400 nonzero entries.

(after computing shared NN)
Allowing nearest neighbours only reduced the number of shared NN from 9133448 to 572722.


Shared nearest-neighbour (connections only) graph (dtype: int16, shape: (11468, 11468)) has 554159 nonzero entries.

Shared nearest-neighbour (number of shared neighbours as weights) graph (dtype: int16, shape: (11468, 11468)) has 554159 nonzero entries.

sNN graph data:
[32 21 26 ... 12 14 14]

---- Ran shared_nn in 0.33 s ----


-- Multiplying sNN connectivity by weights --


shared NN with distance-based weights graph (dtype: float64, shape: (11468, 11468)) has 554159 nonzero entries.
shared NN weighted graph data: [0.0864526  0.08653113 0.08680058 ... 0.24618953 0.25736866 0.27164641]

Converting graph (dtype: float64, shape: (11468, 11468)) has 554159 nonzero entries.
---- Ran csr_to_igraph in 0.08 s ----


Resolution: 0.4
------------------------------

---- Partitioned BANKSY graph ----
modularity: 0.76
7 unique labels:
[0 1 2 3 4 5 6]

---- Ran partition in 2.91 s ----

No annotated labels
```

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

### Import Similarity Metrics

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B006_ascending.ipynb

Imports necessary metrics from scikit-learn for calculating similarity between cluster labels. It also accesses the 'Community' annotation from the adata object.

```python
from sklearn.metrics import adjusted_rand_score as ari, adjusted_mutual_info_score as ami
from sklearn.metrics import matthews_corrcoef as mcc
# See the visualize the communities that we want to detect
adata.obs['Community']
```

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

### Perform Nonspatial Clustering and Visualization

Source: https://github.com/prabhakarlab/banksy_py/blob/main/CODEX_B006_ascending.ipynb

This snippet demonstrates how to set up and run nonspatial clustering using Leiden partitioning and then visualize the results. It requires pre-defined variables like 'adata', 'resolutions', 'max_labels', 'weights_graph', 'coord_keys', and 'unique_regions'. The visualization can be configured to save different plot types.

```python
# Add nonspatial clustering
nonspatial_dict = {"nonspatial" : {0.0: {"adata": concatenate_all([adata.X], 0, adata=adata), } } }

pca_umap(nonspatial_dict, pca_dims = pca_dims, add_umap = True )

from banksy.cluster_methods import run_Leiden_partition

nonspatial_df, max_num_labels = run_Leiden_partition(
    nonspatial_dict,
    resolutions,
    num_nn = 50,
    num_iterations = -1,
    partition_seed = 1234,
    match_labels = True,
    max_labels = max_labels,
)

from banksy.plot_banksy import plot_results

c_map =  'tab20' # specify color map

plot_results(
    nonspatial_df,
    weights_graph,
    c_map,
    match_labels = True,
    coord_keys = coord_keys,
    max_num_labels  =  max_num_labels, 
    save_path = os.path.join(file_path, 'BANKSY-Results'),
    save_fig = False, # Save Spatial Plot Only
    save_fullfig = True, # Save Full Plot
    dataset_name = f"CODEX-{unique_regions}",
    save_labels=True
)
```

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

### Define Sample List and Cluster Targets

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_concatenate_multisample.ipynb

Initializes a list of sample identifiers and a corresponding list for the target number of clusters for each sample. Used for iterating through datasets.

```python
from scanpy import read_10x_h5

samples = [
    "151673","151674","151675","151676",
]

num_clusters_list = [7,7,7,7] # target cluster number
```

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

### Preview Preprocessed Data

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv1_analysis.ipynb

Displays a preview of the preprocessed data, showing various statistical metrics per row. This is useful for a quick overview of the data's structure and quality after preprocessing.

```python
Result:
                  mt  n_cells_by_counts  mean_counts  log1p_mean_counts  \
Row                                                                       
0610005C13Rik  False                  3     0.000117           0.000117   
0610007P14Rik  False                553     0.024696           0.024396   
0610009B22Rik  False                238     0.009784           0.009737   
0610009E02Rik  False                 11     0.000431           0.000430   
0610009L18Rik  False                 67     0.002779           0.002775   
...              ...                ...          ...                ...   
mt-Ts2         False                539     0.021526           0.021297   
mt-Tt          False                173     0.006927           0.006903   
mt-Tv          False               1515     0.063872           0.061915   
n-R5-8s1       False                320     0.012641           0.012562   
n-R5s33        False                  0     0.000000           0.000000   

               pct_dropout_by_counts  total_counts  log1p_total_counts  
Row                                                                     
0610005C13Rik              99.988259           3.0            1.386294  
0610007P14Rik              97.835701         631.0            6.448889  
0610009B22Rik              99.068530         250.0            5.525453  
0610009E02Rik              99.956949          11.0            2.484907  
0610009L18Rik              99.737779          71.0            4.276666  
...                              ...           ...                 ...  
mt-Ts2                     97.890494         550.0            6.311735  
mt-Tt                      99.322923         177.0            5.181784  
mt-Tv                      94.070682        1632.0            7.398174  
n-R5-8s1                   98.747603         323.0            5.780744  
n-R5s33                   100.000000           0.0            0.000000  

[18671 rows x 7 columns]
```

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

### Import Libraries and Warnings

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_harmony_multisample.ipynb

Imports necessary libraries for data analysis, including pandas, numpy, scanpy, and anndata. Suppresses warnings to keep output clean.

```python
import os,csv,re
import pandas as pd
import numpy as np

import math
from scipy.sparse import issparse
import random 
import warnings
warnings.filterwarnings("ignore")

import matplotlib.pyplot as plt

import scanpy as sc
import anndata as ad

import time
```

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

### BANKSY Refinement Output Summary

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

This output provides a summary of the cluster refinement process, including the number of nodes swapped, the ratio of swaps, and the total entropy before and after refinement. It also indicates where the refined plot was saved and the time taken for refinement.

```text
Output:
Refine only once

Number of nodes swapped 26 | ratio: 0.022
Total Entropy: 0.14
Refined once | Total Entropy: 0.14
scaled_gaussian_pc20_nc0.80_r0.90 saved refined plot at W:\\Banksy\\YF_ver\\Banksy_py_CODEX_final\\data\\starmap\\tmp_png\\leiden\\seed1234
Time taken for refinement = 0.01 min
BANKSY runtime = 0.277 mins


```

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

### Display Preprocessed adata with QC Metrics

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv2_analysis.ipynb

Displays the AnnData object after preprocessing and adding quality control metrics. Useful for inspecting the results of QC calculations.

```python
Result:
                  mt  n_cells_by_counts  mean_counts  log1p_mean_counts  \
Row                                                                       \
0610005C13Rik  False                  1     0.000025           0.000025   \
0610007P14Rik  False                679     0.018888           0.018712   \
0610009B22Rik  False                433     0.011672           0.011604   \
0610009E02Rik  False                 33     0.000836           0.000835   \
0610009L18Rik  False                173     0.004608           0.004597   \
...              ...                ...          ...                ...   \
mt-Tv          False                160     0.004051           0.004043   \
n-R5-8s1       False                 23     0.000582           0.000582   \
n-R5s173       False                  1     0.000025           0.000025   \
n-R5s48        False                  1     0.000025           0.000025   \
n-R5s76        False                  1     0.000025           0.000025   \

               pct_dropout_by_counts  total_counts  log1p_total_counts  
Row                                                                     
0610005C13Rik              99.997468           1.0            0.693147  
0610007P14Rik              98.280839         746.0            6.616065  
0610009B22Rik              98.903686         461.0            6.135565  
0610009E02Rik              99.916447          33.0            3.526361  
0610009L18Rik              99.561981         182.0            5.209486  
...                              ...           ...                 ...  
mt-Tv                      99.594896         160.0            5.081404  
n-R5-8s1                   99.941766          23.0            3.178054  
n-R5s173                   99.997468           1.0            0.693147  
n-R5s48                    99.997468           1.0            0.693147  
n-R5s76                    99.997468           1.0            0.693147  

[23096 rows x 7 columns]
```

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

### Plot Marker Gene Sets

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv1_analysis.ipynb

Visualizes marker gene sets using a metagene dataframe, cell coordinates, and RCTD weights. Ensure that the coordinate and weight files are correctly loaded.

```python
from banksy_utils.plot_utils import plot_markergene_sets
import pandas as pd
import os
from scipy.io import mmread

# load weights and locations (need to load locations because filtered seperately)
rctd_coord_filename = "slideseqv1_mc_coords_filtered.csv"
rctd_weights_filename = "slideseqv1_mc_RCTDweights.txt"
rctd_coord = pd.read_csv(os.path.join(file_path, rctd_coord_filename), index_col=0)
rctd_weights = mmread(os.path.join(file_path, rctd_weights_filename)).todense()

plot_markergene_sets(metagene_df,
                     rctd_coord,
                     rctd_weights,
                     coord_keys,
                     save_fig=False)
```

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

### BANKSY Initialization and Matrix Generation

Source: https://github.com/prabhakarlab/banksy_py/blob/main/DLPFC_concatenate_multisample.ipynb

Initializes the BANKSY object with spatial coordinates and generates the BANKSY matrix. This snippet sets up parameters like k_geom, lambda, and m, and includes spatial coordinates in the AnnData object. It also prepares a non-spatial matrix for comparison.

```python
## banksy parameters ##
from banksy.initialize_banksy import initialize_banksy
banksy_dict = {}
coord_keys = ('x_pixel', 'y_pixel', 'coord_xy')
nbr_weight_decay = 'scaled_gaussian'
k_geom = 6
x_coord, y_coord, xy_coord = coord_keys[0], coord_keys[1], coord_keys[2]


from banksy.main import concatenate_all
from banksy.embed_banksy import generate_banksy_matrix
resolutions = [0.60] # clustering resolution for UMAP
pca_dims = [20] # Dimensionality in which PCA reduces to
lambda_list = [0.2] # list of lambda parameters
m = 0
from banksy_utils.umap_pca import pca_umap
from banksy.cluster_methods import run_Leiden_partition
results_df = {}
max_num_labels = {}
from banksy.plot_banksy import plot_results
c_map =  'tab20' # specify color map


### run banksy ###

for sample in samples:
    # Include spatial coordinates information
    raw_y, raw_x = adata_list[sample].obs[y_coord], adata_list[sample].obs[x_coord]
    adata_list[sample].obsm[xy_coord] = np.vstack((adata_list[sample].obs[x_coord].values, adata_list[sample].obs[y_coord].values)).T

    banksy_dict[sample] = initialize_banksy(adata_list[sample],
                                            coord_keys,
                                            k_geom,
                                            nbr_weight_decay=nbr_weight_decay,
                                            max_m=m,
                                            plt_edge_hist= False,
                                            plt_nbr_weights= True,
                                            plt_agf_angles=False,
                                            plt_theta=False
                                            )
    banksy_dict[sample], banksy_matrix = generate_banksy_matrix(adata_list[sample],
                                                                banksy_dict[sample],
                                                                lambda_list,
                                                                max_m=m)
    banksy_dict[sample]["nonspatial"] = {
        # Here we simply append the nonspatial matrix (adata.X) to obtain the nonspatial clustering results
        0.0: {"adata": concatenate_all([adata_list[sample].X], 0, adata=adata_list[sample]), }
    }
    print(banksy_dict[sample]['nonspatial'][0.0]['adata'])
```

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

### Initialize BANKSY Py

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

Initialize the BANKSY Py object with specified parameters and coordinate keys. This function generates spatial graphs and plots related histograms.

```python
banksy_dict = initialize_banksy(
    adata,
    coord_keys,
    k_geom,
    nbr_weight_decay=nbr_weight_decay,
    max_m=max_m,
    plt_edge_hist=True,
    plt_nbr_weights=True,
    plt_agf_angles=False,
    plt_theta=False,
)
```

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

### Run BANKSY with Multiple Parameters

Source: https://github.com/prabhakarlab/banksy_py/blob/main/starmap_analysis.ipynb

Execute the BANKSY analysis pipeline with a comprehensive set of parameters. This function handles data preprocessing, matrix creation, dimensionality reduction, and clustering.

```python
results_df = run_banksy_multiparam(
    adata,
    banksy_dict,
    lambda_list,
    resolutions,
    color_list = spagcn_color,
    max_m = max_m,
    filepath = output_folder,
    key = coord_keys,
    pca_dims = pca_dims,
    annotation_key = annotation_key,
    max_labels = num_clusters,
    cluster_algorithm = cluster_algorithm,
    match_labels = False,
    savefig = False,
    add_nonspatial = False,
    variance_balance = False,
)
```

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

### Display Preprocessed Data with QC Metrics

Source: https://github.com/prabhakarlab/banksy_py/blob/main/slideseqv2_analysis.ipynb

Displays the preprocessed AnnData object, which includes added quality control metrics. This output is useful for initial inspection of the data quality and cell filtering criteria.

```python
import scanpy as sc

# Assuming 'adata' is your AnnData object after QC metric calculation
# Example: adata.obs['n_genes_by_counts'] = (adata.X.sum(axis=1) > 0).sum(axis=1).A1
# Example: adata.obs['total_counts'] = adata.X.sum(axis=1).A1
# Example: adata.var['mt'] = adata.var_names.str.startswith('MT-')
# adata.obs['pct_counts_mt'] = np.sum(adata.X[:, adata.var['mt']], axis=1).A1 / np.sum(adata.X, axis=1).A1

# Display the adata with added qc_metrics
print(adata)
```