In this work we propose a unique approach that, starting from the set of RP disease affected genes, provides a comprehensive landscape of the molecular mechanisms of the disease along with its druggable space. The method establishes a mechanistic disease map as an actionable environment, and employs an explainable machine learning model, “Drug REpurposing using Mechanistic Models of signal transduction and eXplainable Machine Learning” (DRExM³L), to assess the influence of druggable molecules, like drug target proteins, over the disease environment. Our approach merges information from transcriptomics, pathway graphs, biological/clinical, and drug-target interactions databases, to generate an in-depth view of the disease. The novelty of this workflow lies in the integration of multiple data sources, reinforcing interpretability with biological knowledge while reducing the dimensionality of the datasets.
The necessary data to reproduce the results of the manuscript can be downloaded from Zenodo. Note that the workflow will try to automatically download the data after installing the dependencies.
-
The Genotype-Tissue Expression (GTEx) RNA-Seq Data Gene read counts:
- File name: GTEx_Analysis_2017-06-05_v8_RNASeQCv1.1.9_gene_reads.gct
- File path: data/raw
- Source download: https://gtexportal.org/home/datasets
-
The DrugBank 5.1.8 db :
- File name: drugbank-v050108.xml.gz
- File path: data/raw
- Source download: https://go.drugbank.com/releases
-
Disambiguations for ambiguous drug action annotations - DrugBank 5.1.8 db :
- File name: amendments_drugActions_drugbank-v050108.tsv
- File path: data/raw
-
Public databases downloaded through VarfromPDB R package:
- File name: localPDB.tar.gz - static databases April 2020
- Source: https://cran.r-project.org/web/packages/VarfromPDB/VarfromPDB.pdf
-
The Anatomical Therapeutic Chemical (ATC) Classification table:
- File name: ATC.csv
- File path: data/raw
- Source download: https://bioportal.bioontology.org/ontologies/ATC - 2022AB CSV file
-
The Human Phenotype Ontologies database (HPO):
- File name: hp.obo-v1.2-20190906
- File path: data/raw
- Source download: https://hpo.jax.org/app/data/ontology - September 6th, 2019 release
-
The Human Phenotype Ontologies (HPO) annotations linking diseases and phenotypes
- File name: phenotype_annotation20190906.tab
- File path: data/raw
- Source download: https://hpo.jax.org/app/data/annotations - September 6th, 2019 release
-
The Human Phenotype Ontologies (HPO) annotations linking genes and phenotypes
- File name: phenotype_to_genes20191010.txt
- File path: data/raw
- Source download: https://hpo.jax.org/app/data/annotations - September 6th, 2019 release
-
HiPathia's list of physiological KEGG signaling pathways
- File name: physiological_paths.tsv
- File path: data/raw
- Source: http://hipathia.babelomics.org/
-
HiPathia's list of physiological KEGG signaling pathways with GO/Uniprot functional annotations
- File name: physPathsAnnot.tsv
- File path: data/raw
- Source: http://hipathia.babelomics.org/
-
Main actions by drug (manually curated):
- File name: drug_actions_withSimplAction.csv
- File path: data/raw
- Source: https://go.drugbank.com/releases
RP Mechanistic Map here.
The project requires a working conda
and a GNU/Linux x64
system. If conda
is not available,
the workflow will try to install the miniconda
distribution.
Copy example.env
to .env
.
- Set
UPDATE=false
, the default option, to reproduce the manuscript results.
If using a SLURM-based HPC, run the full analysis with:
sbatch run.sbatch
If using bash
, the same script could be used in a non-SLURM system.
We have included a minimal update mechanism to facilitate the reanalysis of the disease:
- Set
UPDATE=true
and we will try to update the databases, use the last version of the Hipathia model, etc.- To update the Drugbank release, copy the
xml
file todata/raw
and changeDRUGBANK_VERSION
in.env
. - To update the GTEx counts, copy the
gct
file todata/raw
and updateGTEX_FNAME
in.env
. - Follow the same procedure with the
hpo.obo
files. - RP experts could update the manually curated files.
- DRExM³L will be automatically updated to the last stable release.
- To update the Drugbank release, copy the
- Set
USE_GPU=N
to use the indicated number of GPUs,N=1, 2, ...
(0 for CPU).
The code of this project is released under the MIT license.
Authors and contributors:
- Marina Esteban-Medina [email protected]
- Kinza Rian [email protected]
- Joaquin Dopazo [email protected]
- Maria Peña-Chilet [email protected]
- Carlos Loucera [email protected]