Skip to content

Code for "Probabilistic pathway based multimodal factor analysis"

Notifications You must be signed in to change notification settings

ratschlab/path-fa

Repository files navigation

Code for "Probabilistic Pathway-based Multimodal Factor Analysis"

The main code for the PathFA method, which is implemented in pathfa/path_fa.py in lines 185 and 420 for the unimodal and multimodal variants. See the docstrings for more details.

Setup and requirements

  • Python >= 3.7 with installed requirements.txt and rpy linked
  • msigdb v7.2 pathways/genesets when re-running data preparation
  • TuPro datasets for reproducing the corresponding experiments

Synthetic Experiments

The synthetic experiments use MSigDB Hallmark pathways and associations prepared in data_synth to simulate example data. To produce all commands that have to be run, execute python runscripts/synthetic_joblist_generate.py > runscripts/synthetic_joblist.sh. Before executing each command, mkdir synthetic_results has to be executed as well. Each line in the synthetic_joblist.sh corresponds to a command that has to be run. For details on the synthetic experiment, see run_synthetic.py. In notebooks/, use the ipynb notebook to produce figures.

TuPro Experiments

The commands are generated with python runscripts/joblist_tupro_generate.py > runscripts/joblist_tupro.sh and each line corresponds to a command that has to be executed. Before executing all commands, make sure to create the directories tupro_results/ovarian and tupro_results/melanoma. Further, make sure to set the variable TUPRO_PATH in pathfa/utils.py to the directory where to load TuPro data from (once it is released). Afterwards, results are aggregated with python scripts/extract_tupro_results.py. In notebooks/, use the ipynb notebook to produce figures.

About

Code for "Probabilistic pathway based multimodal factor analysis"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published