analysis/data/README.md

In krisrs1128/MSLF: Package accompanying "Bootstrap Confidence Regions for Learned Feature Embeddings"

First download the dataset at this link into this directory. Then, unzip it,

tar -zxvf raw_data.tar.gz

You should see the following files / folders,

• spatial_process_models: This contains features from all the trained CNN, VAE, and RCF models across dataset and model parameters in the LCMP spatial process simulation. This is used as input for the bootstrap.Rmd script in analysis/simulation/spatial_process_simulation. Each file is named using the identifiers {model_name}-k{model complexity}-{data fraction}.yaml_{bootstrap number}.tar.gz
• tnbc_models: These are the analogous features for the TNBC spatial proteomics data analysis. Note that only the model type and complexity is varied in this experiment.
• stability_data_sim.tar.gz: This contains the simulated tiles and metadata used to train the models in spatial_process_models. It is not needed in any of the bootstrap scripts in this compendium, but it is used by the python notebooks used to train feature extractors. It was generated using the generate.Rmd file in the spatial process simulation folder.
• stability_data_tnbc.tar.gz: These are the analogous data used for feature learning in the TNBC spatial proteomics dataset. The input tiles for model training were generated using the prepare_mibi.Rmd script in the data analysis folder.

krisrs1128/MSLF documentation built on March 21, 2023, 10:21 a.m.