NETPHIX (NETwork-to-PHenotype association LeveragIng eXlusivity) is a computational tool to identify mutated subnetworks that are associated with a continuous phenotype. Using an integer linear program with properties for mutual exclusivity and interactions among genes, NETPHIX finds an optimal set of genes maximizing the association. For more details, see [1].
- Linux/Mac OS/Windows
- python version 3.6.2/pandas version 0.24.2
- CPLEX version 12.9.0.0 (set PYTHONPATH to the cplex location);
-
install the CPLEX-Python modules usig the script "setup.py" located in "yourCplexhome/python/"
$ python setup.py install -
install python CPLEX interface
$ pip install cplex
-
Run NETPHIX.
usage: run_NETPHIX.py [-h] [-tp TARGET_PERM_FILE] [-np NET_PERM_FILE]
[-ap ALT_PERM_FILE] [-map GENE_MAP_FILE]
[-idx INDEX_NAME] [-sep SEP]
[-restricted RESTRICTED_GENE_FILE] [-max_time MAX_TIME]
[-pool POOL] [-add_file ADD_FILE] [--append]
[--recompute]
alt_file target_file correlation k filter_high
filter_low density norm penalty net_file
solutionfile_name
positional arguments:
alt_file alteration data file (alteration matrix, see
data/Alterataion_data/ directory for an example)
target_file target profile data file (target (phenotype) see
data/Target_data/ directory for an example)
correlation positive, negative, combined (for combined model), or
combined2 (for separate model)
k size of modules
filter_high upperbound mutation frequency of a gene to be included
filter_low lowerbound mutation frequency of a gene to be included
density module density
norm target normalization method, z or zlog
penalty p (penalty) or np (no penalty), penalty to reinforce
mutual exclusivity.
net_file network file
solutionfile_name file name to write the solution
optional arguments:
-h, --help show this help message and exit
-tp TARGET_PERM_FILE, --target_perm_file TARGET_PERM_FILE
file with target permutation results
-np NET_PERM_FILE, --net_perm_file NET_PERM_FILE
file with net permutation results
-ap ALT_PERM_FILE, --alt_perm_file ALT_PERM_FILE
file with alteration table permutation results
-map GENE_MAP_FILE, --gene_map_file GENE_MAP_FILE
file with ENSG ID to gene name mapping
-idx INDEX_NAME, --index_name INDEX_NAME
index name to use in the target file
-sep SEP, --sep SEP separator in file (\t for default)
-restricted RESTRICTED_GENE_FILE, --restricted_gene_file RESTRICTED_GENE_FILE
file containing restricted gene modules. compute the
objective only with genes in the restricted module
-max_time MAX_TIME, --max_time MAX_TIME
time limit for CPLEX
-pool POOL, --pool POOL
gap limit for solution pool
-add_file ADD_FILE, --add_file ADD_FILE
additional mutation file. take OR with the original
mutations
--append add solution to existing file
--recompute recompute solution and write to the existing file
Examples:
Run NETPHIX with drug target 1 (combined, k=3, mutation frequency between 1 and 0.01, network density 0.5, normalization method=z, with penalty)
$ python run_NETPHIX.py data/AlterationsV2_final.txt.gz data/gdsc_auc/Target1.txt combined 3 1 0.01 0.5 z p data/HumanStringNet.txt results/test/result_combined_3_p.txt
Run NETPHIX with drug target 1 (for separate model) (separate model (combined2), k=4, mutation frequency between 1 and 0.01, network density 0.5, normalization method=z, with penalty)
$ python run_NETPHIX.py data/AlterationsV2_final.txt.gz data/gdsc_auc/Target1.txt combined2 4 1 0.01 0.5 z p data/HumanStringNet.txt results/test/result_sep_4_p.txt
Run NETPHIX with drug target 1 (for combined model) (combined, k=3, mutation frequency between 1 and 0.01, network density 0.5, normalization method=z, no penalty)
$ python run_NETPHIX.py data/AlterationsV2_final.txt.gz data/gdsc_auc/Target1.txt combined 3 1 0.01 0.5 z np data/HumanStringNet.txt results/test/result_combined_3_np.txt
Run NETPHIX with drug target 1 and compute p-value with target permutation (assuming target permutation is already performed). See below how to run permutation test (combined, k=3, mutation frequency between 0.25 and 0.01, network density 0.5, normalization method=z)
$ python run_NETPHIX.py data/AlterationsV2_final.txt.gz data/gdsc_auc/Target1.txt combined 3 1 0.01 0.5 z p data/HumanStringNet.txt results/test/result_combined_3_p_with_pv.txt -tp results/test/perm_target_combined_3.txt
Run target (phenotype) permutation. The phenotype is permuted across samples.
usage: target_permute_NETPHIX.py [-h] [-sol ILP_SOL_FILE] [-map GENE_MAP_FILE]
[-idx INDEX_NAME] [-sep SEP]
[-max_time MAX_TIME] [-add_file ADD_FILE]
alt_file target_file correlation k
filter_high filter_low density norm penalty
netfile_name permutations permutefile_name
positional arguments:
alt_file alteration data file (alteration matrix, see
data/Alterataion_data/ directory for an example)
target_file target profile data file (target (phenotype) see
data/Target_data/ directory for an example)
correlation positive, negative, combined (for combined model), or
combined2 (for separate model)
k size of modules
filter_high upperbound mutation frequency of a gene to be included
filter_low lowerbound mutation frequency of a gene to be included
density module density
norm target normalization method, z or zlog
penalty p (penalty) or np (no penalty), penalty to reinforce
mutual exclusivity.
netfile_name network file
permutations num of permutations
permutefile_name file name to write the permutation results
optional arguments:
-h, --help show this help message and exit
-sol ILP_SOL_FILE, --ILP_sol_file ILP_SOL_FILE
file with optimal ILP solution with target permutation
pvalue.
-map GENE_MAP_FILE, --gene_map_file GENE_MAP_FILE
file with ENSG ID to gene name mapping
-idx INDEX_NAME, --index_name INDEX_NAME
index name to use in the target file
-sep SEP, --sep SEP separator in file (\t for default)
-max_time MAX_TIME, --max_time MAX_TIME
time limit for CPLEX
-add_file ADD_FILE, --add_file ADD_FILE
additional mutation file. take OR with the original
mutations
Examples:
Run target permutation 10 times with drug target 1 (use combined model for correlation) (k=3, mutation frequency between 0.25 and 0.01, network density 0.5, normalization method=z, with penalty)
$ python target_permute_NETPHIX.py data/AlterationsV2_final.txt.gz data/gdsc_auc/Target1.txt combined 3 0.25 0.01 0.5 z p data/HumanStringNet.txt 10 results/test/perm_target_combined_3.txt
run_NETPHIX.py
Input:
AlterationsV2_final.txt
target datafiles are at data/gdsc_auc
data/HumanStringNet.txt
Output:
target_permute_NETPHIX.py
Input:
AlterationsV2_final.txt
target datafiles are at data/gdsc_auc
data/HumanStringNet.txt
select_final_modules.ipynb
Input:
netphix files with pv in results/gdsc/
example file for Drug 1
Output:
max_sig_combined_modules_0.05.tsv
Source files
NETPHIX core files
run_NETPHIX.py
- run NETPHIX
target_permute_NETPHIX.py
- run permutation test
select_final_modules.ipynb
- summarize NETPHIX results and pick the final modules
Utility functions
cv_utils.py
dist_utils.py
gen_simulated.py
netphix_utils.py
permute_utils.py
post_analysis.py
select_sig_modules.py
Evalutions
regress_nephix.ipynb
anova_netphix_uncover.ipynb
comp_dist_target_all.ipynb
Data directory
data/
HumanStringNet.txt: interaction network
AlterationsV2_final.txt.gz : gene alteration file
gdsc_auc/Target*.txt : drug response auc files
gdsc_drug_targets.tsv: GDSC drug targets
drug_target_id.txt:
ctrp_auc_processed.txt: CTRP AUC data
merged_uncover_modules_0.05.tsv: UNCOVER modules
Results directory
results/
test/
result_combined_3_p.txt
result_sep_4_p.txt
result_combined_3_np.txt
perm_target_combined_3.txt
result_combined_3_p_with_pv.txt
gdsc/ # netphix candidate modules
depmap_results_$drug_id$_combined_$k$.txt
depmap_results_$drug_id$_combined2_$k$.txt
# netphix final modules
max_sig_combined_modules_0.05.tsv
max_sig_combined_modules_ctrp_cv_0.05.tsv
combination_candidates.tsv
[1] Identifying Drug Sensitivity Subnetworks with NETPHIX. https://www.biorxiv.org/content/10.1101/543876v1?rss=1
- Yoo-Ah Kim (kimy3@ncbi.nlm.nih.gov)