PathwayForte¶
Python package for pathway database benchmarking.
A Python package for benchmarking pathway databases with functional enrichment and prediction methods tasks.
Command Line Interface¶
PathwayForte commands.
pathway_forte¶
Run PathwayForte.
pathway_forte [OPTIONS] COMMAND [ARGS]...
fcs¶
List of FCS Analyses.
pathway_forte fcs [OPTIONS] COMMAND [ARGS]...
gsea¶
Run GSEA on TCGA data.
pathway_forte fcs gsea [OPTIONS]
Options
-
-d,--data<data>¶ Required Name of the cancer dataset from TCGA
- Options
prad | ov | kirc | brca | lihc
-
-p,--permutations<permutations>¶ Number of permutations
- Default
100
ora¶
Perform ORA analysis.
pathway_forte ora [OPTIONS] COMMAND [ARGS]...
hypergeometric¶
Perform one-tailed hypergeometric test enrichment.
pathway_forte ora hypergeometric [OPTIONS]
Options
-
-d,--genesets<genesets>¶ Required Path to GMT file
-
-s,--fold-changes<fold_changes>¶ Required Path to fold changes file
-
--no-threshold¶ Do not apply threshold
-
-o,--output<output>¶ Optional path for output JSON file
prediction¶
List of Prediction Methods.
pathway_forte prediction [OPTIONS] COMMAND [ARGS]...
binary¶
Train elastic net for binary prediction.
pathway_forte prediction binary [OPTIONS]
Options
-
-d,--data<data>¶ Required Name of the cancer dataset from TCGA
- Options
prad | ov | kirc | brca | lihc
-
--outer-cv<outer_cv>¶ Number of splits in outer cross-validation
- Default
10
-
--inner-cv<inner_cv>¶ Number of splits in inner cross-validation
- Default
10
-
-i,--max_iterations<max_iterations>¶ Number of max iterations to converge
- Default
1000
-
--turn-off-warnings¶ Turns off warnings
subtype¶
Train subtype analysis.
pathway_forte prediction subtype [OPTIONS]
Options
-
-d,--ssgsea<ssgsea>¶ Required Path to ssGSEA file
-
-s,--subtypes<subtypes>¶ Required Path to the subtypes file
-
--outer-cv<outer_cv>¶ Number of splits in outer cross-validation
- Default
10
-
--inner-cv<inner_cv>¶ Number of splits in inner cross-validation
- Default
10
-
--chain-pca¶
-
--explained-variance<explained_variance>¶ Explained variance
- Default
0.95
-
--turn-off-warnings¶ Turns off warnings
survival¶
Train survival model.
pathway_forte prediction survival [OPTIONS]
Options
-
-d,--data<data>¶ Required Name of dataset
-
--outer-cv<outer_cv>¶ Number of splits in outer cross-validation
- Default
10
-
--inner-cv<inner_cv>¶ Number of splits in inner cross-validation
- Default
10
-
--turn-off-warnings¶ Turns off warnings
test-stability-prediction¶
Test stability of prediction.
pathway_forte prediction test-stability-prediction [OPTIONS]
Options
-
-s,--ssgsea-scores-path<ssgsea_scores_path>¶ Required ssGSEA scores file
-
-p,--phenotypes-path<phenotypes_path>¶ Required Path to the phenotypes file
-
--outer-cv<outer_cv>¶ Number of splits in outer cross-validation
- Default
10
-
--inner-cv<inner_cv>¶ Number of splits in inner cross-validation
- Default
10
-
-i,--max_iterations<max_iterations>¶ Number of max iterations to converge
- Default
1000
-
--turn-off-warnings¶ Turns off warnings
Pipeline¶
Pipelines from Pathway Forte.
Constants¶
This module contains all the constants used in the PathwayForte repo.
-
pathway_forte.constants.BIO2BEL_DATA_DIR= '/home/docs/.bio2bel/pathwayforte'¶ Cancer Data Sets
-
pathway_forte.constants.make_classifier_results_directory()[source]¶ Ensure that the result folder exists.
-
pathway_forte.constants.MSIG_GSEA= '/home/docs/checkouts/readthedocs.org/user_builds/pathwayforte/checkouts/latest/data/results/gsea/msig'¶ Output files with results for GSEA
-
pathway_forte.constants.make_gsea_export_directories()[source]¶ Ensure that gsea export directories exist.
-
pathway_forte.constants.MSIG_SSGSEA= '/home/docs/checkouts/readthedocs.org/user_builds/pathwayforte/checkouts/latest/data/results/ssgsea/msig'¶ Pickles with results for ssGSEA
-
pathway_forte.constants.make_ssgsea_export_directories()[source]¶ Ensure that gsea export directories exist.
-
pathway_forte.constants.check_gmt_files()[source]¶ Check if GMT files exist and returns GMT files as constant variables.
-
pathway_forte.constants.GENESET_COLUMN_NAMES= {'kegg': 'KEGG Geneset', 'reactome': 'Reactome Geneset', 'wikipathways': 'WikiPathways Geneset'}¶ Columns to read to perform ORA analysis.
Over Representation Methods¶
This module contains the functions to run Over Representation Analysis (ORA).
-
pathway_forte.pathway_enrichment.over_representation.read_fold_change_df(path)[source]¶ Read csv with gene names, fold changes and their p-values.
- Return type
DataFrame
-
pathway_forte.pathway_enrichment.over_representation.filter_p_value(df, p_value=True, cutoff=0.05)[source]¶ Return significantly differentially expressed genes in fold change df.
Functional Class Scoring¶
This module contain the functional class methods implemented in PathwayForte.
Currently this includes GSEA and ssGSEA.
-
pathway_forte.pathway_enrichment.functional_class.create_cls_file(gene_expression_file, normal_sample_file, tumor_sample_file, data)[source]¶ Create categorical (e.g. tumor vs sample) class file format (i.e., .cls) for input into GSEA.
-
pathway_forte.pathway_enrichment.functional_class.run_gsea(gene_exp, gene_set, phenotype_class, permutations=500, output_dir='/home/docs/checkouts/readthedocs.org/user_builds/pathwayforte/checkouts/latest/data/results/gsea')[source]¶ Run GSEA on a given dataset with a given gene set.
-
pathway_forte.pathway_enrichment.functional_class.filter_gsea_results(gsea_results_path, source, kegg_manager=None, reactome_manager=None, wikipathways_manager=None, p_value=None, absolute_nes_filter=None, geneset_set_filter_minimum_size=None, geneset_set_filter_maximum_size=None)[source]¶ Get top and bottom rankings from GSEA results.
- Parameters
gsea_results_path (
str) – path to GSEA results in .tsv file formatsource –
kegg_manager (
Optional[Manager]) – KEGG managerreactome_manager (
Optional[Manager]) – Reactome managerwikipathways_manager (
Optional[Manager]) – WikiPathways managerabsolute_nes_filter (
Optional[float]) – filter by magnitude of normalized enrichment scoresgeneset_set_filter_minimum_size (
Optional[int]) – filter to include a minimum number of genes in a gene setgeneset_set_filter_maximum_size (
Optional[int]) – filter to include a maximum number of genes in a gene set
- Return type
DataFrame- Returns
list of pathways ranked as having the highest and lowest significant enrichment scores
-
pathway_forte.pathway_enrichment.functional_class.merge_statistics(merged_pathways_df, dataset)[source]¶ Get statistics for pathways included in the merged gene sets dataFrame.
These include the proportion of pathways from each of the other databases and the proportion of pathways deriving from 2 or more primary resources
- Parameters
merged_pathways_df (
DataFrame) – dataFrame containing pathways from multiple databases- Returns
statistics of contents in merged dataset
-
pathway_forte.pathway_enrichment.functional_class.rearrange_df_columns(df)[source]¶ Rearrange order of columns.
- Return type
DataFrame
-
pathway_forte.pathway_enrichment.functional_class.get_pathway_names(database, pathway_df, kegg_manager=None, reactome_manager=None, wikipathways_manager=None)[source]¶ Get pathway names from database specific pathway IDs.
-
pathway_forte.pathway_enrichment.functional_class.pathway_names_to_df(filtered_gsea_results_df, all_pathway_ids, source, kegg_manager=None, reactome_manager=None, wikipathways_manager=None)[source]¶ Get pathway names.
- Parameters
- Return type
DataFrame
-
pathway_forte.pathway_enrichment.functional_class.gsea_results_to_filtered_df(dataset, kegg_manager=None, reactome_manager=None, wikipathways_manager=None, p_value=None, absolute_nes_filter=None, geneset_set_filter_minimum_size=None, geneset_set_filter_maximum_size=None)[source]¶ Get filtered GSEA results dataFrames.
-
pathway_forte.pathway_enrichment.functional_class.get_pathways_by_resource(pathways, resource)[source]¶ Return pathways by resource.
- Return type
-
pathway_forte.pathway_enrichment.functional_class.get_analogs_comparison_numbers(kegg_reactome_pathway_df, reactome_wikipathways_pathway_df, wikipathways_kegg_pathway_df, *, pathway_column='pathway_id')[source]¶ Get number of existing versus expected pairwise mappings.
-
pathway_forte.pathway_enrichment.functional_class.get_pairwise_mapping_numbers(kegg_pathway_df, reactome_pathway_df, wikipathways_pathway_df)[source]¶ Get number of existing versus expected pairwise mappings.
-
pathway_forte.pathway_enrichment.functional_class.get_pairwise_mappings(kegg_pathway_df, reactome_pathway_df, wikipathways_pathway_df)[source]¶ Get pairwise mappings.
-
pathway_forte.pathway_enrichment.functional_class.compare_database_results(df_1, resource_1, df_2, resource_2, mapping_dict, check_contradiction=False)[source]¶ Compare pathways in the dataframe from enrichment results to evaluate the concordance in similar pathways.
-
pathway_forte.pathway_enrichment.functional_class.get_matching_pairs(df_1, resource_1, df_2, resource_2, equivalent_mappings_dict)[source]¶ Get equivalent pathways and their direction of change.
-
pathway_forte.pathway_enrichment.functional_class.run_ssgsea(filtered_expression_data, gene_set, output_dir='/home/docs/checkouts/readthedocs.org/user_builds/pathwayforte/checkouts/latest/data/results/ssgsea', processes=1, max_size=3000, min_size=15)[source]¶ Run single sample GSEA (ssGSEA) on filtered gene expression data set.
-
pathway_forte.pathway_enrichment.functional_class.filter_gene_exp_data(expression_data, gmt_file)[source]¶ Filter gene expression data file to include only gene names which are found in the gene set files.
- Parameters
expression_data (
DataFrame) – gene expression values for samplesgmt_file (
str) – .gmt file containing gene sets
- Returns
Filtered gene expression data with genes with no correspondences in gene sets removed
- Return type
pandas.core.frame.DataFrame kegg_xml_parser.py
Pathway Topology Methods¶
This module contain the topology-based topology methods implemented in PathwayForte used R wrappers and are located outside the main Python package in its corresponding R folder at https://github.com/pathwayforte/results/tree/master/R.
Binary Prediction¶
Prediction of binary classes such as tumor vs. normal patients.
Elastic Net regression with nested cross validation module.
This workflow trains an elastic net model for a binary classification task (e.g., tumor vs. normal patients). The training is conducted using a nested cross validation approach (the number of cross validation in both loops can be selected). The model used can be easily changed since most of the models in scikit-learn (the machine learning library used by this package) required the same input.
-
pathway_forte.prediction.binary.ssgsea_nes_to_df(ssgsea_scores_csv, classes_file, removed_random=None)[source]¶ Create dataFrame of Normalized Enrichment Scores (NES) from ssGSEA of TCGA expression data.
-
pathway_forte.prediction.binary.get_l1_ratios()[source]¶ Return a list of values that are used by the elastic net as hyperparameters.
-
pathway_forte.prediction.binary.train_elastic_net_model(x, y, outer_cv_splits, inner_cv_splits, l1_ratio, model_name, max_iter=None, export=True)[source]¶ Train elastic net model via a nested cross validation given expression data.
Uses a defined hyperparameter space for l1_ratio.
- Parameters
x (numpy.array) – 2D matrix of pathway scores and samples
y (list) – class labels of samples
outer_cv_splits (
int) – number of folds for cross validation split in outer loopinner_cv_splits (
int) – number of folds for cross validation split in inner loopl1_ratio (
List[float]) – list of hyper-parameters for l1 and l2 priorsmodel_name (
str) – name of the modelmax_iter (
Optional[int]) – default to 1000 to ensure convergenceexport (
bool) – Export the models usingjoblib
- Return type
- Returns
A list of AUC-ROC scores
Multi-Class Prediction¶
Prediction of multi-class labels such as tumor subtypes.
-
pathway_forte.prediction.multiclass¶
Survival Prediction¶
Prediction of survival based on clinical and pathway patient data.
-
pathway_forte.prediction.survival¶
Utils¶
Complementary methods for prediction analysis.
Utilities for prediction.
-
pathway_forte.prediction.utils.pca_chaining(train, test, n_components)[source]¶ Chain PCA with logistic regression.
- Parameters
train (pandas.core.frame.DataFrame) – Training set to apply dimensionality reduction to
test (pandas.core.series.Series) – Test set to apply dimensionality reduction to
n_components – Amount of variance retained
- Return type
- Returns
array-like, shape (n_samples, n_components)
Mappings Methods¶
Methods related to ComPath mappings.
Function to deal with ComPath mappings.
-
pathway_forte.mappings.get_mapping_dict(df, mapping_type)[source]¶ Create a dictionary with ComPath mappings for each pathway.
-
pathway_forte.mappings.get_equivalent_pairs(df)[source]¶ Get equivalent pairs of pathways from 2 databases.
Installation 
¶
pathway_forte can be installed from PyPI
with the following command in your terminal:
$ python3 -m pip install pathway_forte
The latest code can be installed from GitHub with:
$ python3 -m pip install git+https://github.com/pathwayforte/pathway-forte.git
For developers, the code can be installed with:
$ git clone https://github.com/pathwayforte/pathway-forte.git
$ cd pathway-forte
$ python3 -m pip install -e .
Main Commands¶
The table below lists the main commands of PathwayForte.
Command |
Action |
|---|---|
datasets |
Lists of Cancer Datasets |
export |
Export Gene Sets using ComPath |
ora |
List of ORA Analyses |
fcs |
List of FCS Analyses |
prediction |
List of Prediction Methods |
Functional Enrichment Methods¶
ora. Lists Over-Representation Analyses (e.g., one-tailed hyper-geometric test).
fcs. Lists Functional Class Score Analyses such as GSEA and ssGSEA using GSEAPy.
Prediction Methods¶
pathway_forte enables three classification methods (i.e., binary classification, training SVMs for
multi-classification tasks, or survival analysis) using individualized pathway activity scores. The scores can be
calculated from any pathway with a variety of tools (see 1) using any pathway database that enables to export its
gene sets.
binary. Trains an elastic net model for a binary classification task (e.g., tumor vs. normal patients). The training is conducted using a nested cross validation approach (the number of cross validation in both loops can be selected). The model used can be easily changed since most of the models in scikit-learn (the machine learning library used by this package) required the same input.
subtype. Trains a SVM model for a multi-class classification task (e.g., predict tumor subtypes). The training is conducted using a nested cross validation approach (the number of cross validation in both loops can be selected). Similarly as the previous classification task, other models can quickly be implemented.
survival. Trains a Cox’s proportional hazard’s model with elastic net penalty. The training is conducted using a nested cross validation approach with a grid search in the inner loop. This analysis requires pathway activity scores, patient classes and lifetime patient information.
Other¶
References¶
- 1
Lim, S., et al. (2018). Comprehensive and critical evaluation of individualized pathway activity measurement tools on pan-cancer data. Briefings in bioinformatics, bby125.
- 2
Domingo-Fernández, D., et al. (2018). ComPath: An ecosystem for exploring, analyzing, and curating mappings across pathway databases. npj Syst Biol Appl., 4(1):43.
- 3
Weinstein, J. N., et al. (2013). The cancer genome atlas pan-cancer analysis project. Nature genetics, 45(10), 1113.