Multi-Dendrix Pipeline

The simplest way to use Multi-Dendrix is to run the entire Multi-Dendrix pipeline. We have provided a program to run a customizable version of the Multi-Dendrix pipeline in the multi_dendrix_pipeline.py script. The analysis performed in the pipeline, the parameters for customizing the analysis, and documentation of the program itself are presented below.

Outline

The Multi-Dendrix pipeline (implemented in multi_dendrix_pipeline.run()) consists of the following steps:

1. Runs Multi-Dendrix on input mutation data for a range of number t of gene sets and maximum gene set size kmax. This yields a set of collections of gene sets (see multi_dendrix_pipeline.batch_multi_dendrix()). The motivation for running Multi-Dendrix on a range of parameter settings is that while we expect genes in the same functional pathway to have approximate exclusivity in their mutations, we do not know the sizes or the number of these pathways. By running on a range of parameter sizes, we hope to identify these true values (detailed in the next step).
2. Analyze the collections identified in 1. for the gene sets that appear together consistently across parameter choices (see multi_dendrix.core_modules.extract()).
3. Evaluate each collection for statistical signficance using a matrix permutation test. See multi_dendrix_pipeline.run_matrix_permutation_test() and multi_dendrix.evaluate.matrix.matrix_permutation_test() for details.
4. Evaluate each collection for enrichment of protein-protein interactions on a protein-protein interaction (PPI) network. See multi_dendrix_pipeline.run_network_permutation_test() and multi_dendrix.evaluate.network.direct_interactions_test() for details.
5. Analyze all genes (or mutation classes) for (sub)type-specific mutations (if subtypes are known). See multi_dendrix.subtypes.subtype_analysis() for details.
6. Output the results of the analysis as both text and HTML files.

At this time, the pipeline does not include any functions for preprocessing mutation data.

Arguments

The Multi-Dendrix pipeline accepts the following parameters. You may also be interested in the Input File Formats accepted by Multi-Dendrix.

Name	Type	Description	Flag(s)	Required	Default	Example
Output directory	string	Directory name to store the text and HTML output.	-o --output_dir	Yes.	N/A	"output"
Verbose	boolean	Flag whether to output progress or not.	-v --verbose	No.	False	N/A
kmin	integer	Minimum gene set size.	-k_min --min_gene_set_size	Yes.	N/A	2
kmax	integer	Maximum gene set size.	-k_max --max_gene_set_size	Yes.	N/A	3
tmin	integer	Minimum number of gene sets.	-t_min --min_num_gene_sets	Yes.	N/A	2
tmin	integer	Maximum number of gene sets.	-t_max --max_num_gene_sets	Yes.	N/A	3
Data name	string	Name of data (for use in naming output files).	-n --db_name	Yes.	N/A	"BRCA"
Mutation matrix file	string	Location of input mutation matrix.	-m --mutation_matrix	Yes.	N/A	"BRCA.m2"
Mutation frequency cutoff	integer	Minimum number of patients a gene must be mutated in to be included in the mutation data.	-c --cutoff	No.	0	2
Patient whitelist	string	See description of white- and blacklists.	-p --patient_whitelist	No.	None	"BRCA.lst"
Patient blacklist	string	See description of white- and blacklists.	-bp --patient_blacklist	No.	None	"BRCA.blst"
Gene whitelist	string	See description of white- and blacklists.	-g --gene_whitelist	No.	None	"BRCA.glst"
Gene blacklist	string	See description of white- and blacklists.	-bg --gene_blacklist	No.	None	"fishy_genes.glst"
α	float	Changes tradeoff between coverage and coverage overlap in the weight function used by Multi-Dendrix.	-a --alpha	No.	1.0	3.0
Δ	integer	The number of overlaps allowed between gene sets in a single collection output by Multi-Dendrix.	-d --delta	No.	0	1
λ	integer	The number of gene sets a gene is allowed to be a member in a single collection output by Multi-Dendrix.	-l --lmbda	No.	1	2
Stability threshold for core modules	integer	The number of gene sets a pair of genes must be a member of together to be grouped into the same core module.	--stability_threshold	No.	2	1
Perform subtype analysis	boolean	Flag whether the pipeline should include subtype-specific mutation analysis.	--subtypes	No.	False	N/A
Significance level for subtype-specific mutations	float	The maximum Bonferonni-corrected p-value for reported subtype-specific mutations.	--subtype_sig_threshold	No.	0.05	0.01
Perform network test	boolean	Flag whether to perform the network permutation test.	--network_test	No.	False	N/A
Input PPI network	string	File location of an input PPI network (formatted as a NetworkX edge list).	-ppi --network_edgelist	No.	None	"iref_edgelist"
Number of permuted networks	integer	Number of permuted PPI networks to create when performing the direct interactions test. Only used when a directory location of premuted networks is not provided.	--num_permuted_networks	No.	5	1000
Permuted networks directory	string	Directory location of permuted network edge lists.	--permuted_networks_dir	No.	None	"permuted_iref/"
Average pairwise distance flag	boolean	Flag whether to perform the average pairwise distance test instead of the direct interactions test.	--distance	No.	False	N/A
Q	integer	Parameter to control the number of edge swaps when permuting PPI networks. For graph G=(V, E), Q * |E| edge swaps are performed.	--Q	No.	100	5
Perform statistical significance test	boolean	Flag whether to perform the statistical significance test.	--weight_test	No.	False	N/A
Number of permuted matrices	integer	Number of permuted matrices to create when calculating statistical significance. Only used when a directory location of premuted matrices is not provided.	--num_permuted_matrices	No.	5	100
Permuted matrices directory	string	Directory location of permuted matrices.	--permuted_matrices_dir	No.	None	"permuted_BRCA/"

Script

run(args)	Runs the whole Multi-Dendrix Pipeline for the given command-line arguments.
batch_multi_dendrix(args)	Runs Multi-Dendrix for each parameter setting on the input
run_network_permutation_test(args, ...)	Runs the direct interactions or average pairwise distance test on each of the collections and the core_modules.
run_matrix_permutation_test(args, ...)	Runs the direct interactions or average pairwise distance test on each of the collections and the core_modules.