A heterogeneous and integrated knowledge graph for the exploration of neglected tropical diseases
- About the project
- How to run the web app locally?
- Structure of the repository
- Credits
- Further details
- Contact
Neglected tropical diseases (NTDs) are a heterogeneous group of 20 bacterial, viral, parasitic, and fungal conditions that generally occur in developing tropical countries in the Americas, Africa, and Asia. NTDs mainly affect poor populations that do not have access to safe water, sanitation, and high-quality healthcare. Because of the severe effects of NTDs (i.e., they can cause long-lasting disabilities), they reinforce the cycle of poverty in vulnerable communities.
Currently, there are several independent databases that contain information about proteins, metabolic pathways, and drugs involved in NTDs, but no integrated databases with all the information. This unified resource could enable the systematic exploration of all the components of the NTDs, contributing to the research of potential therapies for these diseases.
The NTDs2RDF project aimed to create a knowledge graph (KG) of genes, proteins, metabolic pathways, gene ontologies, single nucleotide variants, drugs, and other relevant data for three NTDs (Chagas disease, leishmaniasis, and African trypanosomiasis), integrating all the information in a single data structure that can be explored through a query interface implemented with a Streamlit web application. This software provides a user-friendly platform to extract information from the KG using SPARQL queries.
An overview of the workflow of this project is shown in the graphical abstract below.
Figure 1. Graphical abstract of the NTDs2RDF workflow.
The project represents an initial step towards the creation of a heterogeneous database for different NTDs with several potential applications in advancing the understanding of NTDs biology and providing insights that cannot be obtained through alternative resources.
This project was divided into various modules for data collection, data cleaning, programmatic creation of the KG, and the development of a web application for querying the RDF graph. The software architecture of all the parts of the NTDs2RDF project is presented in the figure below.
Figure 2. Software architecture of the NTDs2RDF project. The workflow was divided into three main stages with some modules (blue
rectangles) and components (green rectangles).
The NTDs2RDF KG has ten classes and 33.892 triplets. The Biolink model was the main source to define classes and predicates of the RDF graph because it has many human-readable and domain-specific predicates for Biology, and its entities also follow a machine-readable format and they are integrated with other ontologies using semantic mappings, which makes the KGs interoperable.
To assure that all the entities of the KG have standardized and stable identifiers, the Bioregistry metaregistry was used to assign the URIs and CURIEs. The figure below shows the meta-graph of the NTDs2RDF KG with all the classes and predicates from the Biolink Model and other standard ontologies.
Figure 3. Meta-graph of the NTDs2RDF KG. The classes are represented as blue ovals, the literals as green rounded rectangles, the
references to external databases as orange rounded rectangles, and the provenance for the classes as yellow rounded rectangles.
To guarantee the accessibility and findability of the NTDs2RDF KG, a multi-page Streamlit web application was developed. This software was divided into three pages: Meta-graph with all the classes and predicates from the Biolink model and other standard ontologies, examples of SPARQL queries to extract relevant information from the graph, and the query interface. The web application is available online at the link below, and it can also be run locally by following the instructions in the next section.
I used Pipenv to create a Python virtual environment, which allows the management of python libraries and their dependencies. Each Pipenv virtual environment has a Pipfile with the names and versions of libraries installed in the virtual environment, and a Pipfile.lock, a JSON file that contains versions of libraries and their dependencies.
To create a Python virtual environment with libraries and dependencies required for this project, you should clone this GitHub repository, open a terminal, move to the folder containing this repository, and run the following commands:
# Install pipenv
$ pip install pipenv
# Create the Python virtual environment
$ pipenv install
# Activate the Python virtual environment
$ pipenv shellYou can find a detailed guide on how to use pipenv here.
Alternatively, you can create a conda virtual environment with the required libraries using the requirements.txt file.
After installing the libraries, you can run the streamlit app locally with the command below:
$ streamlit run 🏠_Home.pyThe main files and directories of this repository are:
| File | Description |
|---|---|
| RDF_graph_building.ipynb | Jupyter notebook to integrate the data and create the RDF graph |
| NTDs2RDF_report.pdf | Full research report and biological/technical discussion |
| 🏠_Home.py | Script for the home page of the streamlit web application |
| sparql_queries_NTDs_RDF_examples.txt | Examples of SPARQL queries to retrive information from the RDF graph |
| requirements.txt | File with names of the libraries required for the streamlit web application |
| Pipfile | File with names and versions of libraries installed in the virtual environment |
| Pipfile.lock | Json file that contains versions of libraries and their dependencies |
| style.css | css file to customize style features of the web application |
| Data/ | Raw csv files and RDF graph |
| pages/ | Python scripts for the pages of the streamlit web application |
| Data_processing/ | R scripts to collect the data |
| img/ | images and gifs |
-
Developed by Sebastián Ayala Ruano. I created this app for a project of a course about knowledge graphs from the MSc in Systems Biology at Maastricht University.
-
Part of the code for this project was inspired by the Medium blogpost and GitHub repo from Edoardo Bianchi about this topic.
The PDF Manuscript contains more background information on NTDs, a detailed description of the methodology, creation of the RDF graph, and the web application, as well as a discussion of the limitations and potential applications of this project.
If you have comments or suggestions about this project, you can open an issue in this repository.