Cricca

This repository contains the source code for Parameterized algorithms for identifying gene co-expression modules via weighted clique decomposition.

Requirements

numpy, pandas, networkx, scipy, gurobi

To Run Your Own Data

• From /cricca direcotory, run

    python src,main.py --graph_filename [your_fname.txt] --algorithm [either 'wecp', 'ipart', or 'lp'] --parameter [integer parameter value]
  

• For information on other input arguments, run

    python src/main.py --help
  

To Recreate Paper Experiments

1. Graph Generation

Steps to generate the same corpus used in the paper experiments (see paper for additional graph generation details.):

1. Download Transcription Factor Base Data: TF Data

2. Open Catalog.xls, save sheet 'TFactS_sign_less_version2' as 'gene_TF_dat.csv' in generator_dat directory.

3. Download Latent Variable Base Data (Note file is 84GB): LV Data

4. Place the files named 'multiplier_model_z.tsv' and 'multiplier_model_summary.tsv' in generator_dat directory.

5. From /cricca directory, run

   python src/graph_gen/gen_acda21_corpus.py
   

Both TF and LV graphs will now exist in the src/data/pre_preprocessing directory.

2. Preprocessing

To preprocess each graph, from /cricca directory, run:

python src/exp_preprocess.py -f 0 -l 19

-f/--first_seed and -l/--last_seed give the seeds of desired graphs to preprocess.

3. Kernelization

To kernelize each preprocessed graph with k [2-11] (after pre-processing), from /cricca directory, run:

python src/exp_runkernel.py -f 0 -l 19

4. Matrix Decomposition

To run each matrix decomposition algorithm including wecp (original algorithm by Feldmann et al. 2020), ipart (integer partitioning based), and lp (linear programming based) algorithms, from /cricca directory, run:

python src/exp_runbsd.py -f 0 -l 19

Note: timeout code in function run_bsd() in file exp_runbsd.py only works on unix systems (e.g. signal.signal/alarm(....) function calls).

5. CSV Creation

To combine all data saved in pickle files (after running exp_runbsd.py) into a single csv file, from /cricca directory, run:

python src/tocsv.py

Additional Tests

To run the memory tests comparing wecp/ipart/lp algorithms, from /cricca directory, run:

python src/exp_runmemtests.py