Note

This page was generated from Slices Alignment with Paste.ipynb. Interactive online version: . Some tutorial content may look better in light mode.

Slices Alignment with PASTE

This notebook demonstrates the process of Aligning spatial transcriptomics data. This is done in the following two steps:

1. Align spatial transcriptomics data from a set of multi-slices;

2. Reduce the amount of data by sampling to improve alignment speed.

Here we modified a published method PASTE, which utilized Fused Gromov-Wasserstein Optimal Transport (FGW-OT) algorithm, to leverages both gene expression similarity and spatial distances between spots to align and integrate spatial transcriptomics data.

Reference: Ron Zeira, Max Land, Benjamin J. Raphael. Alignment and Integration of Spatial Transcriptomics Data. bioRxiv, 2021.03.16.435604. doi: https://doi.org/10.1101/2021.03.16.435604

import warnings
warnings.filterwarnings('ignore')

import spateo as st

2023-07-24 17:44:11.458267: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

Load the data

# cellbin data
cellbin_slices = st.sample_data.drosophila(filename="E7-9h_cellbin_h5ad.zip")[4:8]
cellbin_slices

[AnnData object with n_obs × n_vars = 1745 × 9057
     obs: 'area', 'slices'
     uns: '__type', 'spatial'
     obsm: 'bbox', 'contour', 'spatial',
 AnnData object with n_obs × n_vars = 1769 × 9158
     obs: 'area', 'slices'
     uns: '__type', 'spatial'
     obsm: 'bbox', 'contour', 'spatial',
 AnnData object with n_obs × n_vars = 1244 × 8992
     obs: 'area', 'slices'
     uns: '__type', 'spatial'
     obsm: 'bbox', 'contour', 'spatial',
 AnnData object with n_obs × n_vars = 1715 × 9145
     obs: 'area', 'slices'
     uns: '__type', 'spatial'
     obsm: 'bbox', 'contour', 'spatial']

Align spatial transcriptomics data from a set of multi-slices

Visualize slices based on raw coordinates

st.pl.multi_slices(
    slices=cellbin_slices,
    slices_key="slices",
    spatial_key="spatial",
    point_size=10,
    ncols=2,
)

Slices alignment

aligned_slices, pis = st.align.paste_align(
    models=[slice.copy() for slice in cellbin_slices],
    spatial_key="spatial",
    key_added="align_spatial",
    device="0" # or device="cpu"
)

|-----> [Models alignment] in progress: 33.3333%|-----> Filtered all samples for common genes. There are 8334 common genes.
|-----> [Models alignment] in progress: 66.6667%|-----> Filtered all samples for common genes. There are 8257 common genes.
|-----> [Models alignment] in progress: 100.0000%|-----> Filtered all samples for common genes. There are 8251 common genes.
|-----> [Models alignment] in progress: 100.0000%
|-----> [Models alignment] finished [13.8075s]

Visualize slices based on aligned coordinates

st.pl.multi_slices(
    slices=aligned_slices.copy(),
    slices_key="slices",
    spatial_key="align_spatial",
    point_size=10,
    ncols=2,
)

Reduce the amount of data by down-sampling to improve alignment speed

Slices alignment

aligned_slices, downsampling_slices, pis = st.align.paste_align_ref(
    models=[slice.copy() for slice in cellbin_slices],
    models_ref=None,
    n_sampling=500,
    sampling_method="trn",
    spatial_key="spatial",
    key_added="align_spatial",
    numItermax=400,
    device="0"
)

|-----> [Running TRN] in progress: 100.0000%|-----> [Running TRN] completed [19.5275s]
|-----> [Running TRN] in progress: 100.0000%|-----> [Running TRN] completed [19.2842s]
|-----> [Running TRN] in progress: 100.0000%|-----> [Running TRN] completed [19.7409s]
|-----> [Running TRN] in progress: 100.0000%|-----> [Running TRN] completed [19.7497s]
|-----> [Models alignment] in progress: 33.3333%|-----> Filtered all samples for common genes. There are 8334 common genes.
|-----> [Models alignment] in progress: 66.6667%|-----> Filtered all samples for common genes. There are 8257 common genes.
|-----> [Models alignment] in progress: 100.0000%|-----> Filtered all samples for common genes. There are 8251 common genes.
|-----> [Models alignment] in progress: 100.0000%
|-----> [Models alignment] finished [0.5662s]

Visualize downsampling slices based on aligned coordinates

st.pl.multi_slices(
    slices=downsampling_slices.copy(),
    slices_key="slices",
    spatial_key="align_spatial",
    point_size=10,
    ncols=2,
)

Visualize slices based on aligned coordinates

st.pl.multi_slices(
    slices=aligned_slices.copy(),
    slices_key="slices",
    spatial_key="align_spatial",
    point_size=10,
    ncols=2,
)