Skip to content

uhlerlab/STACI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

96 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

STACI

Please note that the "useA" parameter in the training notebook need to be set to True, if using spatial locations of cells/spots is desired. Setting useA=False trains the autoencoder with only gene expression.

Package versions:

  • Python 3.10.4
  • pytorch 1.9.1
  • scanpy 1.9.1
  • scipy 1.7.3
  • numpy 1.22.4
  • scikit-learn 1.1.2
  • matplotlib-base 3.5.2
  • matplotlib 3.5.2
  • seaborn 0.11.2
  • pandas 1.4.3
  • umap-learn 0.5.3
  • anndata 0.8.0
  • mnnpy 0.1.9.5

All neural network models and image processing functions are stored in "image" and "gae".

Training autoencoders and plotting the results can be done in the notebooks listed below, following the instructions provided in the notebooks: 0. Script for computing the adjacency matrices: getXA_starmap.ipynb. A user provided matrix can also be used directly without this step.

  1. Training script for graph convolutional autoencoder: train_gae_starmap_multisamples.ipynb

  2. Training script for joint latent space: train_jointGAEcnn_starmap_multisamples.ipynb

  3. Plotting graph autoencoder's latent space: plotEmbedding_Starmap.ipynb

  4. Translation from joint latent space to gene expression: translation_jointCNNgae2Starmap_final.ipynb

  5. Nuclei segmentation: segment3D_gpu.ipynb

  6. Training script for the regression of plaque size from joint latent space: train_regrsFromJoint_starmap_multisamplesMixed.ipynb

  7. Analyzing the results of regression of plaque size from joint latent space: plotRegrsFromJoint_starmap_3Dseg.ipynb

The following notebooks are for generating the validation results we used in our paper and are not necessary for training or analyzing new models:

Validation with 10x Visium data of mouse brain coronal sections: notebooks labeled with "10xADFFPE"

Testing overparameterization on scRNA-seq data to remove batch effects: notebooks labeled with "scrnaseq"

Validation on the four held-out STARmap samples: notebooks labeled with "newdata"

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages