Skip to content

transfer learning for survival recurrent neural networks.

License

Notifications You must be signed in to change notification settings

insilica/TransferSRNN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

40 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TransferSRNN

Here is an explanation of the files in this repo.

Discrete Survival Demo.ipynb - Notebook that shows whole process of building and running survival models. Can be run in new environment after pip installing requirements.txt. Takes a couple hours to run. In the end, it generates a plot similar to lossdiff_boxplot2.svg

lossdiff_boxplot.svg - Plot of difference between loss of respective fold and disease of baseline model and transfer model

binding/baseline_losses.csv - Loss values for each fold and disease for baseline models

binding/transfer_losses.csv - Loss values for each fold and disease for transfer models

Feature importance table generated by iterating across all features and randomly shuffling one feature while holding all others constant. The more that the loss increases when shuffling a particular feature, the more important we can assume that feature to be.

binding/global_feature_importance.csv - Feature importance table for global model.

binding/ind_disease_transfer_feature_importance.csv - Feature importance table for each specific disease

binding/shuffled_hallmark_disease_importance.csv - Hallmark importance generated by only shuffling those set of genes simultaneously and holding other genes constant (and finding diff in loss with no shuffling at all) for specific diseases

binding/shuffled_hallmark_global_importance.csv - Hallmark importance for global model

requirements.txt - Python dependencies to be installed in new environment

nnet_survival.py - Python file containing survival model functions

genes_pubmedcounts_glioma.csv - Table of gene names and number of PubMed articles regarding the gene. Articles restricted to those listed when searching for glioma MeSH term.

genes_pubmedcounts_prostate.csv - Table of gene names and number of PubMed articles regarding the gene. Articles restricted to those listed when searching for prostatic neoplasm MeSH term.

To run Discrete Survival Demo.ipynb, run the following command to install dependencies in your environment:

pip install -r requirements.txt

Next, in Discrete Survival Demo.ipynb's 2nd cell, cd to TransferSRNN directory which you git cloned.

Now you can run all the cells

About

transfer learning for survival recurrent neural networks.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published