This repository is part of team Heartbreak's project for the course Data and Visual Analytics in Spring 2019 at Georgia Institute of Technology

The purpose of this project is to identify protein-protein interactions in biomedical literature

The project has two parts

1. Backend
2. Frontend

The backend is responsible for providing an REST API that takes as input a paper in pdf or txt format and outputs a JSON that describes the genes and their interactions. When the backend starts, it trains the classifier using the transformed data in the datasets/training directory

The frontend is the user-interface made in React. It uses Material-UI to render a modern looking frontend for the project. For visualizing the graph D3.js is used.

There are two ways to run this project

1. Using Docker (the easy way)
2. Using Python and NodeJs as usual

• Windows 10 Pro, Linux or MacOS
• Docker

1. Open terminal in the project root folder
2. Run docker-compose up -d

• Python 2.7
• NodeJs (Latest LTS version)
• Java 8 Runtime (JRE)

1. Go to the repository directory
2. Install python dependencies. pip install requirements.txt
3. Download NLTK packages. python -m nltk.downloader punkt averaged_perceptron_tagger universal_tagset
4. Run python app.py

1. Go the directory <repository-root>/web-ui
2. Install npm dependencies. npm install
3. Run npm start

1. If running through Docker. Navigate to localhost:4000 else localhost:3000
2. Click the button "Choose File". Only pdf and txt file formats are supported.
3. Select the file PIIS1097276506003376.pdf in the papers folder in the repository root
4. Click "ANALYZE FILE"
5. Wait for the graph
6. Drag and drop the nodes to move the nodes

• Custom made dataset of gene interaction sentences
• BioC-BioGRID
• reviewed-home-sapien-genes.tab from UniProt

• NLTK
• scikit-learn
• PyBioC
• React
• D3
• Material-UI