Spatial domain refinement¶

This tutorial demonstrates how to perform spatial domain refinement using SpaHDmap, focusing on a specific region of interest within your spatial transcriptomics data.

We will use the IHC stained mouse brain dataset as an example, refining a region corresponding to a specific spatial domain.

1. Import Necessary Libraries¶

[2]:

import torch
import numpy as np
import scanpy as sc

import SpaHDmap as hdmap

/home/qk/anaconda3/lib/python3.11/importlib/__init__.py:126: FutureWarning: The legacy Dask DataFrame implementation is deprecated and will be removed in a future version. Set the configuration option `dataframe.query-planning` to `True` or None to enable the new Dask Dataframe implementation and silence this warning.
  return _bootstrap._gcd_import(name[level:], package, level)

2. Set Parameters and Paths¶

Set the parameters and file paths for analysis.

[3]:

rank = 20
seed = 123
verbose = True

np.random.seed(seed)
torch.manual_seed(seed)

root_path = '../experiments/'
project = 'MBC01'
results_path = f'{root_path}/{project}/Results_Rank{rank}/'
results_path_refine = f'{root_path}/{project}/Refine_Emb2/'

4. Run SpaHDmap Analysis for refined section¶

We define a function to generate a mask for the refined section, which covers the spots of interest.

[6]:

def get_mask(section):
    mask = np.zeros_like(section.mask, dtype=np.bool_)
    for (row, col) in section.spot_coord:
        row, col = round(row - section.row_range[0]), round(col - section.col_range[0])
        row_range = np.arange(max(row - 2*section.radius, 0), min(row + 2*section.radius + 1, mask.shape[0]))
        col_range = np.arange(max(col - 2*section.radius, 0), min(col + 2*section.radius + 1, mask.shape[1]))
        mask[np.ix_(row_range, col_range)] = 1

    return mask

mouse_cortex_refine.mask = get_mask(mouse_cortex_refine)

The mask is applied to the refined section to focus on the spots of interest.

[7]:

import matplotlib.pyplot as plt
plt.imshow(mouse_cortex_refine.mask*255, cmap='gray')
plt.show()

../_images/tutorials_Domain-refinement_13_0.png

[8]:

mapper = hdmap.Mapper(mouse_cortex_refine, results_path=results_path_refine, rank=10, verbose=True)

mapper.run_SpaHDmap()

*** Preparing the tissue splits and creating pseudo spots... ***
*** Single section detected. Using its 10000 genes. ***
*** The split size is set to 256 pixels. ***
For section mouse_cortex, divide the tissue into 824 sub-tissues, and create 2500 pseudo spots.
*** Using GPU ***
Step 1: Run NMF
*** Performing NMF... ***
*** Visualizing and saving the embeddings of NMF... ***
Step 2: Pre-train the SpaHDmap model
*** Pre-trained model found at ../experiments//MBC01/Refine_Emb2//models//pretrained_model.pth, loading... ***
Step 3: Train the GCN model
*** Performing GCN... ***
*** Training GCN for mouse_cortex... ***
[Iter: 200 / 5000], Loss: 0.017658, Learning rate: 4.985215e-03
[Iter: 400 / 5000], Loss: 0.010385, Learning rate: 4.941093e-03
[Iter: 600 / 5000], Loss: 0.008443, Learning rate: 4.868331e-03
[Iter: 800 / 5000], Loss: 0.007181, Learning rate: 4.768075e-03
[Iter: 1000 / 5000], Loss: 0.006268, Learning rate: 4.641907e-03
[Iter: 1200 / 5000], Loss: 0.005731, Learning rate: 4.491816e-03
[Iter: 1400 / 5000], Loss: 0.005271, Learning rate: 4.320170e-03
[Iter: 1600 / 5000], Loss: 0.004972, Learning rate: 4.129675e-03
[Iter: 1800 / 5000], Loss: 0.004755, Learning rate: 3.923336e-03
[Iter: 2000 / 5000], Loss: 0.004506, Learning rate: 3.704407e-03
[Iter: 2200 / 5000], Loss: 0.004469, Learning rate: 3.476340e-03
[Iter: 2400 / 5000], Loss: 0.004159, Learning rate: 3.242732e-03
[Iter: 2600 / 5000], Loss: 0.004052, Learning rate: 3.007268e-03
[Iter: 2800 / 5000], Loss: 0.003961, Learning rate: 2.773660e-03
[Iter: 3000 / 5000], Loss: 0.003808, Learning rate: 2.545593e-03
[Iter: 3200 / 5000], Loss: 0.003665, Learning rate: 2.326664e-03
[Iter: 3400 / 5000], Loss: 0.003471, Learning rate: 2.120325e-03
[Iter: 3600 / 5000], Loss: 0.003364, Learning rate: 1.929830e-03
[Iter: 3800 / 5000], Loss: 0.003285, Learning rate: 1.758184e-03
[Iter: 4000 / 5000], Loss: 0.003210, Learning rate: 1.608093e-03
[Iter: 4200 / 5000], Loss: 0.003162, Learning rate: 1.481925e-03
[Iter: 4400 / 5000], Loss: 0.003108, Learning rate: 1.381669e-03
[Iter: 4600 / 5000], Loss: 0.003074, Learning rate: 1.308907e-03
[Iter: 4800 / 5000], Loss: 0.003039, Learning rate: 1.264785e-03
[Iter: 5000 / 5000], Loss: 0.003032, Learning rate: 1.250000e-03
*** Visualizing and saving the embeddings of GCN... ***
Step 4: Run Voronoi Diagram
Step 5: Train the SpaHDmap model
*** Trained model found at ../experiments//MBC01/Refine_Emb2//models//trained_model.pth, loading... ***
*** Extracting SpaHDmap scores for mouse_cortex... ***
*** Visualizing and saving the embeddings of SpaHDmap... ***

Now, we can visualize the refined SpaHDmap scores, where the index 1 corresponds to the Thalamus, sensory-motor cortex related region (DORsm) and index 7 corresponds to the Thalamus, polymodal association cortex related region (DORpm)

[9]:

mapper.visualize(mouse_cortex_refine, use_score='SpaHDmap', index=1)

*** Visualizing and saving the embeddings of SpaHDmap... ***

../_images/tutorials_Domain-refinement_16_1.png

[11]:

mapper.visualize(mouse_cortex_refine, use_score='SpaHDmap', index=7)

*** Visualizing and saving the embeddings of SpaHDmap... ***

../_images/tutorials_Domain-refinement_17_1.png

[21]:

mapper.cluster(use_score='SpaHDmap', resolution=0.2, show=True)

*** Performing clustering using SpaHDmap scores... ***
Found 3 clusters for section mouse_cortex
*** Visualizing clustering results for SpaHDmap... ***

../_images/tutorials_Domain-refinement_18_1.png

Spatial domain refinement¶

1. Import Necessary Libraries¶

2. Set Parameters and Paths¶

3. Load stData and Prepare for Refinement¶

4. Run SpaHDmap Analysis for refined section¶