BioPhi is an open-source antibody design platform. It features methods for automated antibody humanization (Sapiens), humanness evaluation (OASis) and an interface for computer-assisted antibody sequence design.
BioPhi is available at: http://biophi.dichlab.org
Learn more in the BioPhi, Sapiens and OASis in our pre-print:
Prihoda, D., Maamary, J., Waight, A., Juan, V., Fayadat-Dilman, L., Svozil, D., & Bitton, D. A. (2021). BioPhi: A platform for antibody design, humanization and humanness evaluation based on natural antibody repertoires and deep learning. BioRxiv, 2021.08.08.455394. https://doi.org/10.1101/2021.08.08.455394
The data and notebooks supporting the analysis are found in: https://github.com/Merck/BioPhi-2021-publication
BioPhi is an open and extensible platform, contributions are welcome.
If you have ideas about what to improve or which tools could be integrated, please submit any feature requests using the Issues tab.
If you don't want to use the public BioPhi server, you can run BioPhi on your own machine.
To run BioPhi with OASis humanness evaluation locally, you will need to download and unzip the OASis database file (22GB uncompressed).
# Download database file
wget https://zenodo.org/record/5164685/files/OASis_9mers_v1.db.gz
# Unzip
gunzip OASis_9mers_v1.db.gz
You can install BioPhi using Conda or one of the alternatives (Miniconda, Miniforge).
Set up Bioconda and Conda-Forge channels:
conda config --add channels bioconda
conda config --add channels conda-forge
Install BioPhi using:
# Recommended: Create a separate BioPhi environment
conda create -n biophi python=3.8
conda activate biophi
# Install BioPhi
conda install biophi
# Set up path to OASis database (downloaded and unzipped)
export OASIS_DB_PATH=/path/to/downloaded/OASis_9mers_v1.db
# Run simplified BioPhi server (not for live deployment!)
biophi web
Note: This is simplified usage for local use only. See Deploying your own BioPhi server section below to learn about deploying BioPhi properly on a server.
BioPhi also provides a command-line interface that enables bulk processing.
See more
# Get humanized FASTA
# Expected input: Both chains of each antibody should have the same ID
# with an optional _VL/_VH or _HC/_LC suffix
biophi sapiens mabs.fa --fasta-only --output humanized.fa
# Run full humanization & humanness evaluation pipeline
biophi sapiens mabs.fa \
--oasis-db path/to/downloaded/OASis_9mers_v1.db \
--output humanized/
# Get the Sapiens probability matrix (score of each residue at each position)
biophi sapiens mabs.fa --scores-only --output scores.csv
# Get mean Sapiens score (one score for each sequence)
biophi sapiens mabs.fa --mean-score-only --output scores.csv
# Get OASis humanness evaluation
biophi oasis mabs.fa \
--oasis-db path/to/downloaded/OASis_9mers_v1.db \
--output oasis.xlsx
BioPhi is composed of three services that need to be running at the same time:
web
: Flask web server that handles both the frontend and the backend of the web applicationcelery
: Asynchronous worker service(s) that process long-running tasksredis
: In-memory database for storing celery queue tasks and results
Running through Docker Compose is easiest in terms of setup, but web server autoreload is not supported, so you will have to restart the services after each code update.
See more
See https://docs.docker.com/get-docker/
# Run using Makefile
make docker-build
# or directly using
docker-compose build
# Run using Makefile
make docker-run
# or directly using
docker-compose up
To build and run, you can use:
# Run using Makefile
make docker-build docker-run
# or directly using
docker-compose up --build
After your code is updated, you will need to stop the services, run build and start again. See the next section for info on running locally with flask auto-reload.
Running each service locally using Conda will enable flask auto-reload, which is useful if you are going back and forth between your IDE and the browser.
See more
Install Conda or one of the alternatives (Miniconda, Miniforge)
Install and run Redis server. On Mac, you can install Redis using Brew.
# Install dependencies using the provided Makefile
make env
# Or directly using
conda env create -n biophi -f environment.yml
conda activate biophi
pip install -e . --no-deps
You will have to run each service in a separate terminal (Use Cmd+T to open a new tab):
# Run Redis server (this depends on your installation, the server might already be running)
redis-server
# In a separate terminal, run celery worker queue
make celery
# In a separate terminal, run flask web server
make web
See the provided
After your code is updated, the flask web service should refresh automatically. However, the celery service needs to be stopped and started manually, so you will need to do that if you update code that is executed from the workers.
You can deploy your own internal BioPhi server. You will need to run the three separate services - the flask web server, the celery worker and the redis database.
This will depend on your platform and your cloud provider, the easiest deployment is using Docker Compose through the provided docker-compose.yml file.
For 🐧 Ubuntu deployment, feel free to copy the deployment configs used on the public university server: lich-uct/biophi.dichlab.org
BioPhi is based on antibody repertoires from the Observed Antibody Space:
Kovaltsuk, A., Leem, J., Kelm, S., Snowden, J., Deane, C. M., & Krawczyk, K. (2018). Observed Antibody Space: A Resource for Data Mining Next-Generation Sequencing of Antibody Repertoires. The Journal of Immunology, 201(8), 2502–2509. https://doi.org/10.4049/jimmunol.1800708
Antibody numbering is performed using ANARCI:
Dunbar, J., & Deane, C. M. (2016). ANARCI: Antigen receptor numbering and receptor classification. Bioinformatics, 32(2), 298–300. https://doi.org/10.1093/bioinformatics/btv552