simulations

Scripts and visualization notebooks used to generate simulations on real data (keyboard dataset: Qiita 232) and benchmark rclr transform.

• simulations.ipynb
• intro_fig.ipynb
• figures
  • figure2.png (figure generated in simulations.ipynb)
• scripts
  • cluster_build.py (script runs simulation building on keyboard data in cluster_models)
  • cluster_results.py (runs simulation results in cluster_models)
• cluster_models (simulations and results)
  • keyboard.biom.gz (Real data used to fit simulations)
  • keyboard.txt.gz (Real meta-data used to fit simulations)
  • simulation_subsampled_noisy.csv.gz (simulation with noise/sparsity generated from cluster_build.py)
  • base_model_keyboard_table.csv.gz (simulation without noise/sparsity noise generated from cluster_build.py)
  • ordination (benchmark ordinations from cluster_results.py)
  • results.csv.gz (results kl-div from cluster_results.py)

case_studies (Requires Qiime2, run here on qiime2-2018.6)

Scripts and visualization notebooks used to benchmark rclr-RPCA against Bray-Curtis and Weighted UniFrac for two case studies:

• Case Study 1: The Sponge Microbiome Project (Qiita 10793)
• Case Study 2: Sleep Apnea (Qiita 10422)
• figures
  • figure3.png (benchmarking results on subsamples from subsample_results.tar.gz)
  • figure4.png (use case example from sub_sample max samples in both cases)
• case_studies.ipynb (visualization notebook)
• data.tar.gz (data used acquired from Qiita 10/01/2019)
• subsample_results.tar.gz (results from running all scripts)
• scripts
  • subsample.py (generate subsamples from the data.tar.gz dir)
  • distances.py (run distances for each subsample RPCA, WU, and BC)
  • sub_results.py (return PERMANOVA and KNN results)
• sub_sample
  • biom_tables_Sleep_Apnea (all sleep apnea subsamples [70,184] 10-fold generaterd by subsample.py)
  • biom_tables_Sponges (all sponge subsamples [70,248] 10-fold generated by subsample.py)