config.yaml

# Input Settings: initialize base input folder names,
# and dataset collections
input_settings:

    # Base input directory
    input_dir: "inputs"

    # "datasets" denotes a list of experiments to run, each with the folling specified (a * indicates required options):
    #   *net_version: name to give this network. Will be used to organize output files 
    #
    #   net_files: list of one or more network files, 
    #       or a .mat file containing multiple networks.
    #       multi_net must be set to True if multiple networks are to be combined.
    #       the path should be from <input_dir>/<net_version>
    #
    #   string_net_files: a list of full STRING files, meaning all 16, tab-delimited columns are present
    #       the path should be from <input_dir>/<net_version>
    #
    #   *exp_name: the name to give the experiment to run
    #
    #   *pos_neg_file: a 3 column, tab-delimited file with two lines per term.
    #       Col1: term, Col2: 1 or -1 denoting positive or negative, and a comma-separated list of proteins
    #
    #   only_functions_file: a file containing the GO term IDs to run.
    #
    #   taxon_file: a tab-delimited file indicating the NCBI taxononmy ID (2nd col) for each gene/UniProt ID (1st col).
    #       Necessary to perform Leave-One-Species-Out (LOSO) evaluation.
    #
    #   only_taxon_file: a file containing the taxon IDs to evaluate. Also used to get the names of the species
    #
    #   pos_neg_file_eval: same as the pos_neg_file, only these annotations are used to evaluate.
    #       Some use cases: temporal holdout, or recovering annotations of another evidence code
    #
    #   
    datasets:
        - net_version: "2018_06-seq-sim-e0_1"
          net_files: ["2018_06-seq-sim-e0_1-net.txt"]
          #string_net_files: [""]
          exp_name: "expc-bp"
          pos_neg_file: "pos-neg/expc/expc-bp.tsv"
          #only_functions_file: ""
          # if passing in multiple networks or one or more 'string_net_files',
          # set multi_net to True and use these settings
          multi_net: False
          net_settings:
              # method to combine the networks. either swsn, gmw, or unweighted
              weight_method: 'swsn'
              # if full STRING files are given, 'string_nets' and 'string_cutoff' are required
              ## either core, nontransferred, all, or a comma-separated list of the string networks, or a combination of the above. Default: all
              #string_nets: 'core'
              ## cutoff of the STRING scores. Default: 150
              #string_cutoff: 400
              # If a .mat file is given, 'net_names_file' and 'node_ids_file' are required.
              net_names_file: ""
              node_ids_file: ""



# the outputs will follow this structure:
# outputs/<net_version>/<exp_name>/<alg_name>/output_files
output_settings:
    output_dir: "outputs/"

# Algorithms to run along with their settings.
# All combinations of parameters will be run. 
algs: 
    # local has no parameters
    local:
        should_run: [False]

    fastsinksource:
        should_run: [True]
        alpha: [0.95]
        eps: [0]
        max_iters: [10]
    
    # sinksource-bounds
    # TODO finish refactoring 
    sinksource-bounds:
        should_run: [False]
        rank_all: [False]
        rank_pos_neg: [False]
        compare_ranks: [False]
        alpha: [0.95]
        max_iters: [10]

    # genemania
    genemania:
        should_run: [False]
        tol: [1e-05]

    # birgrank
    birgrank:
        should_run: [False]
        # path to the OBO file from the input_dir
        obo_file: ["inputs/goa/2017_09/2017-09-26-go.obo"]
        alpha: [.5]
        lambda: [.5]
        mu: [.5]
        theta: [.5]
        eps: [1e-04]
        max_iters: [1000]
    
    # aptrank
    aptrank:
        should_run: [False]
        # path to the OBO file from the input_dir
        obo_file: ["inputs/goa/2017_09/2017-09-26-go.obo"]
        lambda: [.5]
        k: [8]
        s: [5]
        t: [0.5]
        diff_type: ["twoway"]

    #fastsinksourceplus:
    #    alpha: 0.9
    #    eps: 0.0001
    #    max_iters: 1000
    #    # this option isn't needed if alpha is < 1
    #    sinksourceplus_lambda: 0

# Evaluation settings for CV or LOSO can be set here 
# TODO these should probably be unique per dataset
eval_settings:

    ### LOSO parameters
    # Don't leave out annotations when running the algorithms.
    # Can only be used in conjunction with a pos_neg_file_eval
    keep_ann: False
    # Minimum number of annotations for each term in the "left-out" species to test. Default: 10
    num_test_cutoff: 10
    # if --pos-neg-file-eval is given and --keep-ann is False, only evaluate terms that have at least 2% of annotations.
    # Useful to speed-up processing for term-based algorithms
    eval_goterms_with_left_out_only: False
    # Postfix to append to output file. Useful if running multiple in parallel. TODO figure out how to automatically combine the multiple output files
    postfix: ""