Network Medicine Framework to Predict Health Impact of Dietary Polyphenols

Code and input files for the proximity-based analysis in the "Predicting the Health Impact of Dietary Polyphenols Using a Network Medicine Framework" manuscript.

Instructions
Network Proximity
Implicit Associations
Gene Set Enrichment Analysis
Data Files
Citation

Instructions

Clone repository

git clone https://github.com/italodovalle/polyphenols.git

Install requirements

cd polyphenols/
pip install -r requirements.txt

Network Proximity

run_proximity.py

usage: run_proximity.py [-h] -i INTERACTOME -d DISEASEGENES -t TARGETS
                        [-c CORES] [-r NRANDOM] [-o OUTFOLDER]
                        [-sn SOURCE_NODE] [-tn TARGET_NODE] [-lcc LCC]
                        [-sep SEPARATOR] [-test TEST_RUN] [-chem CHEMICAL_COL]
                        [-prot PROTEIN_COL] [-tmp TMP_DIR]

Calculate Network Proximity

optional arguments:
  -h, --help          show this help message and exit
  -i INTERACTOME      interactome file
  -d DISEASEGENES     disease genes file
  -t TARGETS          chemical protein interactions
  -c CORES            number of cores to use, default = 6
  -r NRANDOM          n random permutations, default = 1000
  -o OUTFOLDER        outfolder
  -sn SOURCE_NODE     name source node Interactome, default = proteinA
  -tn TARGET_NODE     name target node Interactome, default = proteinB
  -lcc LCC            Consider only Interactome LCC, default = True
  -sep SEPARATOR      separator Interactome, default = ","
  -test TEST_RUN      test run
  -chem CHEMICAL_COL  chemical column in chemical-protein interaction file
  -prot PROTEIN_COL   chemical column in chemical-protein interaction file
  -tmp TMP_DIR        temporary folder

Examples

Example

./run_proximity.py -i data/HumanInteractome_v2017.csv -d data/PeroxisomalDisorders.csv -t data/Genistein.csv -r 1000 -sn EntrezA -tn EntrezB -chem chemical -prot entrez_id

run time: 30 seconds

Entire dataset

./run_proximity.py -i data/HumanInteractome_v2017.csv -d data/GenesDisease.csv -t data/PolyphenolProteinInteractions.csv -c 6 -r 1000 -tmp tmp/ -sn EntrezA -tn EntrezB -test True -chem chemical -prot entrez_id

run time: it can take several days to calculate the proximity for the entire dataset.

Implicit Associations

Consider the example of the disease 'Cardiovascular diseases' (C14.XXX):
- The following chemicals have the 'therapeutic' label in CTD: Genistein, Kaempferol, Resveratrol, Daidzein
Now, if you look at the specific diseases that are under 'Cardiovascular diseases' in the MeSH hierarchy we find the following chemicals with 'therapeutic' labels in CTD:
- cardiomyopathies C14.280.238 (Narigenin, Apigenin)
- cardiomyopathy, hypertrophic C14.280.484.150.070.160 (Quercetin)
- heart diseases C14.280 (-)-Epigallocatechin 3-O-gallate
- etc
Therefore, we can consider, for example, that the association 'narigenin' and 'cardiovascular diseases' is implicit
Therefore, we mapped all explicit associations in the MeSH hierarchy (4816 diseases) to expand explicit to implicit associations
MeSH mapping to the different branches:
- all branches were considered
Considering only diseases under the branch C (Diseases). Some phenotypes in CTD might be under the branch F (Psycology and Psychiatry)
MeSH Supplementary Concept Data ID were manually mapped
Input
- data/mtrees2018.bin
- data/ctd_disease_chemical_phenolexplorer_therapeutic.csv
Output:
- data/ctd_polyphenols_implicit_explicit.csv

expand_associations.py

Usage

usage: expand_associations.py [-h] -m MESH -i INFILE [-o OUTFILE]

Expand associations in MeSH tree

optional arguments:
  -h, --help  show this help message and exit
  -m MESH     MeSH File
  -i INFILE   association table
  -o OUTFILE  outfile

Example:

python expand_associations.py -m data/mtrees2018.bin -i data/ctd_disease_chemical_phenolexplorer_therapeutic.csv -o data/ctd_polyphenols_implicit_explicit.csv

Gene Set Enrichment Analysis

run_gsea.py

Usage

usage: run_gsea.py [-h] -indir INDIR -i INTERACTOME -d DISEASEGENES -t TARGETS
                   [-c CORES] [-r NRANDOM] [-o OUTFOLDER] [-sn SOURCE_NODE]
                   [-tn TARGET_NODE] [-lcc LCC] [-sep SEPARATOR]
                   [-chem CHEMICAL_COL] [-prot PROTEIN_COL] [-tmp TMP_DIR]
                   [-cell CELL_LINE]

Calculate Enrichment Score

optional arguments:
  -h, --help          show this help message and exit
  -indir INDIR        folder with .gct files
  -i INTERACTOME      interactome file
  -d DISEASEGENES     disease genes file
  -t TARGETS          chemical protein interactions
  -c CORES            number of cores to use, default = 6
  -r NRANDOM          n random permutations, default = 1000
  -o OUTFOLDER        outfolder
  -sn SOURCE_NODE     name source node Interactome, default = proteinA
  -tn TARGET_NODE     name target node Interactome, default = proteinB
  -lcc LCC            Consider only Interactome LCC, default = True
  -sep SEPARATOR      separator Interactome, default = ","
  -chem CHEMICAL_COL  chemical column in chemical-protein interaction file
  -prot PROTEIN_COL   chemical column in chemical-protein interaction file
  -tmp TMP_DIR        temporary folder
  -cell CELL_LINE     restrict to cell line, default = MCF7

Example

./run_gsea.py -indir data/connectivity_map/ -i data/HumanInteractome_v2017.csv -d data/PeroxisomalDisorders.csv -t data/Genistein.csv -r 10 -sn EntrezA -tn EntrezB -chem chemical -prot entrez_id

Results written in out_gsea
Run time: It take hours/days for the entire dataset

Data Files

The data folder contains:

data/HumanInteractome_v2017.csv: Human interactome (see manuscript for details)
data/GenesDisease.csv: Disease-gene associations for MeSH disease terms curated in Menche et al. (2015, Science)
data/PolyphenolProteinInteractions.csv: Polyphenol protein targets obtained from the STITCH database

Note

Polyphenol protein interactions were obtained from STITCH (http://stitch.embl.de/download/protein_chemical.links.detailed.v5.0/9606.protein_chemical.links.detailed.v5.0.tsv.gz). We used PubChem CIDS to represent chemicals and we mapped protein ENSEMBL ids to ENTREZ IDs as the final representation for proteins. We only considered interactions in which the experimental column was > 0.
data/connectivity_map: Perturbation profiles obtained in the Connectivity Map database (https://clue.io/) for selected polyphenols
data/mtrees2018.bin: MeSH hierarchy
ctd_disease_chemical_phenolexplorer_therapeutic.csv: all disease-polyphenol associations obtained by combining data from the databases CTD and PhenolExplorer
ctd_polyphenols_implicit_explicit.csv: output file of the code expand_associations.py
Genistein.csv: Genistein targets (for tests)
PeroxisomalDisorders.csv: Peroxisomal Disorders disease genes (for tests)

Citation

do Valle, I. F. et al. Network medicine framework shows that proximity of polyphenol targets and disease proteins predicts therapeutic effects of polyphenols. Nat. Food 2, 143–155 (2021).

italodovalle / polyphenols