Skip to content

Files & folders

SD edited this page Jan 21, 2020 · 1 revision

List of pipeline folders

  • data: input datasets (networks in edge_lists, extra node features optionally (node_attributes), and the labeled protein coding genes based on the disease gene list (train_node_labels) and optionally the test set genes (test_node_labels).

  • parameters: parametrization of the models.

  • src: core code of Tiresias. Includes all models, bagging, LOOCV, artifacts folder creation (intermediate files), mlruns visualization folder creation (AUC of the cumulative distribution curve, ranking of unlabeled nodes, assessment metrics such as Spearman’s coefficient).

List of pipeline files

  • README: instructions for Tiresias installation and use.

  • config.yml: file to be user-update with the relevant system resources, input files paths, and the methods selection.

  • Makefile: all available steps of the run, the pipeline of Tiresias.

  • environment.yml: dependencies of the run.

  • data/node_attributes: optional extra node features file- already provided or user-provided.

  • data/train_node_labels: file with all protein coding genes and their corresponding label according to the disease gene list. A tab-delimited file with 2 columns: node and label. Node is the index number of the indexed protein coding gene list (), and label is 0 for genes not included in the disease gene list, and 1 for these that are.

  • data/test_node_labels: optional file similar to the above (train_node_labels) but where the labels are 0 for genes not included in the test set gene list, and 1 for these that are.

  • data/edge_lists/layer*.tsv: network examples. Each network is in a tab-delimited file with 3 columns: src, dst, weight. Unweighted networks need to be weighted (1 as weight of each edge).

  • parameters/features: file with random walks and skip gram (embeddings) parameters.

  • parameters/models_validation: file with all model parameters for the runs of the validation step of the pipeline, the run on the disease gene list.

  • parameters/models_test: file with all model parameters for the runs for the optional test step of the pipeline, the run on the test set gene list.

List of model configurations

Model name Description
direct neighbors label propagation method- no learning
label_spreading label propagation method- no learning
rwr random walk with restart- no learning
rwr_m random walk with restart for multilayer networks- no learning
bagging_logistic_regression node2vec (embeddings) coupled with logistic regression
bagging_logistic_regression_with_attributes node2vec (embeddings) coupled with logistic regression for learning, with extra gene features
bagging_mlp node2vec (embeddings) coupled with multilayer perceptron (MLP) for learning
bagging_mlp_with_attributes node2vec (embeddings) coupled with multilayer perceptron (MLP) for learning, with extra gene features
bagging_gcn graph convolutional networks
bagging_gcn_with_attributes graph convolutional networks, with extra gene features
bagging_rgcn relational graph convolutional networks for weighted networks
bagging_rgcn_with_attributes relational graph convolutional networks for weighted networks, with extra gene features

Clone this wiki locally