search

greenelab / tybalt

Training and evaluating a variational autoencoder for pan-cancer gene expression data
BSD 3-Clause "New" or "Revised" License
162 stars 62 forks source link

Updating code and visualization for ADAGE tied weights #126

Closed gwaybio closed 6 years ago

gwaybio commented 6 years ago

This PR updates code and visualizations for:

1) Short Latent Z Space Sweep 2) WGCNA simulation

There shows 79 files updated, but I don't think the PR will be too tough to review. I believe it will be easiest to review commits individually, in two parts. I will outline the changes below

Part I. Short Latent Z Space Sweep Updates

  1. df017a4 - Updating scripts/num_components_paramsweep.py to accept untied weights argument
  2. e3e6abc - Updating scripts/adage_pancancer.py to add tied weights command line argument
  3. c088f52 - Updating scripts/viz/param_sweep_latent_space_viz.R to accommodate tied weights results
  4. 4861dcc - Updating all short latent z space sweep figures

Part II. WGCNA Simulation Updates

  1. d0befa4 - Updating run_simulation.py to accept adage tied weights option and updating scripts/util/aggregate_simulation_results.py to aggregate results based on command arguments
  2. 3f3a7fd - Updating scripts/viz/plot_simulation.R to accept results with tied weights ADAGE and adding all updated figures
  3. aacb427 - Update simulation_results.md with new ADAGE interpretation

I want to also link this PR to #125 - the results are getting unruly! It is becoming difficult to separate out analyses and this PR is a good example. The next PR will be focused on reorganization and address this.

It is also important to link to #127 - the current ADAGE results do not pass the sniff test. These visualization and implementation updates are mostly independent of possible required updates pending investigation

gwaybio commented 6 years ago

Thanks for comments @jaclyn-taroni and @danich1 :smile: - I believe this is ready again for review