Pipeline for running and evaluating algorithms for gene/protein annotation prediction, e.g., to the Gene Ontology (GO) and to the Human Phenotype Ontology(HPO).

• Required Python packages: networkx, numpy, scipy, pandas, sklearn, pyyaml, tqdm, rpy2
• Required R packages: PPROC

conda create -n ann-pred python=3.7 r=3.6 --file requirements.txt
conda activate ann-pred

To install the R packages:

R -e "install.packages('https://cran.r-project.org/src/contrib/PRROC_1.3.1.tar.gz', type = 'source')"

If you are unable to install the the R package for computing the AUPRC and AUROC, the code will use sklearn instead, which is not as accurate in some cases.

The script will automatically generate predictions from each of the given methods with should_run: [True] in the config file. The default number of predictions stored is 10. To write more, use either the --num-pred-to-write or --factor-pred-to-write options (see python run_eval_algs.py --help). For example:

python run_eval_algs.py  --config config.yaml --num-pred-to-write -1

The relevant options are below. See python run_eval_algs.py --help for more details.

• cross_validation_folds
  • Number of folds to use for cross validation. Specifying this parameter will also run CV
• cv_seed
  • Can be used to specify the seed to use when generating the CV splits.

Example:

python run_eval_algs.py  --config config.yaml --cross-validation-folds 5 --only-eval

After CV has finished, to visualize the results, use the plot.py script. For example:

python plot.py --config config.yaml --box --measure fmax

If you use FastSinkSource or other methods in this package, please cite:

Jeffrey N. Law, Shiv D. Kale, and T. M. Murali. Accurate and Efficient Gene Function Prediction using a Multi-Bacterial Network, Bioinformatics (2020).