The GUI for RaptGen developed with React and FastAPI

How to Launch

Preliminaries

Please check if the Docker is installed. like

$ docker -v
Docker version 20.10.21, build baeda1f

Procedure

1. Open your terminal. If you would like to run this application on a remote server, use SSH with port-forwarding.

   $ ssh -L 3000:localhost:3000 username@hostname.com

   Otherwise, skip this step.
2. Clone this repository wherever you want, then go into RaptGen-UI directory.

   $ git clone https://github.com/hmdlab/RaptGen-UI.git
   $ cd RaptGen-UI

3. Export your UID and GID environmental variables with the following command (needed for the worker container to work successfully.)

   $ export UID GID

4. Build and run containers with docker-compose. If you have GPU devices which supports CUDA, run with docker-compose.gpu.yml file.

   $ docker compose -f docker-compose.gpu.yml up -d

   Otherwise, you need to assign docker-compose.prod.yml file.

   $ docker compose -f docker-compose.prod.yml up -d

5. Please wait before all the containers are ready. This may take a few minutes. Even if Docker says they are ready, it may take some extra time for the frontend container to be working.
6. Access http://localhost:3000 with your favorite internet browser.
7. If you would like to stop the containers, please type the following command. This stops containers and all data will be retained in db container.

   $ docker compose stop

   If you send down command, all data will be lost (containers are removed.)

Usage

For now, four application is available. They are Viewer, VAE Trainer, GMM Trainer, and Bayesian Optimization. For more information, please refer to the following links.

Viewer

Visualize the latent map of the HT-SELEX data.

You can encode a single nucleotide sequence or batch sequences from fasta file. However decoding from a batch coordinates file is not supported. Downloading is also supported. You can select which cluster to download.

VAE Trainer

Train a VAE model on HT-SELEX data.

GMM Trainer

Train a GMM model on latent space of HT-SELEX data.

Bayesian Optimization

Optimize aptamers using Bayesian Optimization.