Configure your run

1) Edit the configuration file config.yml with:

a. Number of CPUs, GPUs, and memory in GB available to run the tool.
b. Your input data with the absolute or relative path to the files below:

• edge lists: networks
• node attributes: extra gene features
• train_node_labels: labeled protein coding gene list
• test_node_labels: OPTIONAL, same as above with labels from the test set. If no test_node_labels file is indicated it will take a random sample from the train_node_labels file. -test_fraction: fraction of samples from train_node_labels taken to generate the test_node_labels is such is not indicated.float from 0 to 1.

c. Your choice of models to run: True, for enabling False, for disabling.

2) If needed, edit the Makefile:

In order to use part of the provided pipeline, remove the part you do not want to use from the last line. For example, if you have already run the data preparation step:

`.PHONY: pipeline`
`pipeline:       ## Run the whole pipeline.`
        `$(MAKE) data random-walks embeddings validation test predict`

becomes

`.PHONY: pipeline`
`pipeline:       ## Run the whole pipeline minus the data preparation.`
        `$(MAKE) random-walks embeddings validation test predict`

3) Hyperparameters search

Hyperparameters can be customized for each model by modifying the features.json, models_validation.json, models_test.json, in the parameters folder. Models can be optimized based on the available computational resources in terms of run time and based on each targeted disease. The size of the search space depends on the computational and time resources of the user.