georgeamccarthy / protein_search

The neural search engine for proteins.

protein_search

Neural search through protein sequences using the ProtBert model and the Jina AI framework.

App demo:

protein_search.mp4

• ProtBert https://huggingface.co/Rostlab/prot_bert
• Jina AI https://jina.ai
• 3D protein models http://3dmol.org

Setting up the environment

First, clone the repository with git,

git clone https://github.com/georgeamccarthy/protein_search/ # Cloning
cd protein_search # Changing directory

✔️ I have Docker

If you're familiar with Docker, you can simply run make docker (assuming you're running Linux).

The above command will,

1. Create the container for the frontend, installs dependencies, starts the Streamlit application
2. Create the container for the backend, installs dependencies, starts the Jina application
3. Provide you with links as logs to access the two containers

Visually, you should see something like,

From there on, you should be able to visit the Streamlit frontend, and enter your protein relatd query.

Some notes before you use this route,

1. Docker takes a few moments to build the wheel for the dependencies, so the pip step in each of the containers my last as long as 1-2 minutes.
2. The torch dependency in backend/requirements.txt is 831.1 MBs large at the time of writing. Unless you get red colored logs, everything is fine and just taking time to be installed for torch
3. This project uses the Rostbert/prot_bert pre-trained model from HuggingFace which is 1.68 GBs in size.

The great news is that you will need to install these dependencies and build the images only once. Docker will cache all of the layers and steps, and caching for the pre-trained model has been integrated.

Some more functionalites provided are,

• To stop the logs from docker, press Ctrl^C
• For resuming, run make up
• To remove the containers from the background, run make remove
• To build the containers again, run make docker

As for introducing new changes, both the containers do not need to be restarted to do so.

❌ I don't use Docker

For each of the folders frontend, and backend, run the following commands

• Making a new venv virtual environment,

cd folder_to_go_into/ # `folder_to_go_into` is either `frontend` or `backend`
python3 -m venv env
source venv/bin/activate

• Installing dependencies

pip install -r requirements.txt

If in backend, run python3 src/app.py

Open a new terminal, head back into the frontend folder, repeat venv creation and dependency installation, and run streamlit run app.py.

Formatting, linting and testing

Refer to the Makefile for the specific commands

To format code following the black standard

$ make format

Code linting with flake8

$ make lint

Testing

$ make test

Testing with coverage analysis

$ make coverage

Format, test and coverage

$ make build