Sapiens is a human antibody language model based on BERT.

Last update: Nov 20, 2022

Overview

Sapiens: Human antibody language model

    ____              _                
   / ___|  __ _ _ __ (_) ___ _ __  ___ 
   \___ \ / _` | '_ \| |/ _ \ '_ \/ __|
    ___| | |_| | |_| | |  __/ | | \__ \
   |____/ \__,_|  __/|_|\___|_| |_|___/
               |_|

Sapiens is a human antibody language model based on BERT.

Learn more in the Sapiens, OASis and BioPhi in our publication:

David Prihoda, Jad Maamary, Andrew Waight, Veronica Juan, Laurence Fayadat-Dilman, Daniel Svozil & Danny A. Bitton (2022) BioPhi: A platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning, mAbs, 14:1, DOI: https://doi.org/10.1080/19420862.2021.2020203

For more information about BioPhi, see the BioPhi repository

Features

Infilling missing residues in human antibody sequences
Suggesting mutations (in frameworks as well as CDRs)
Creating vector representations (embeddings) of residues or sequences

Usage

Install Sapiens using pip:

# Recommended: Create dedicated conda environment
conda create -n sapiens python=3.8
conda activate sapiens
# Install Sapiens
pip install sapiens

❗️ Python 3.7 or 3.8 is currently required due to fairseq bug in Python 3.9 and above: pytorch/fairseq#3535

Antibody sequence infilling

Positions marked with * or X will be infilled with the most likely human residues, given the rest of the sequence

import sapiens

best = sapiens.predict_masked(
    '**QLV*SGVEVKKPGASVKVSCKASGYTFTNYYMYWVRQAPGQGLEWMGGINPSNGGTNFNEKFKNRVTLTTDSSTTTAYMELKSLQFDDTAVYYCARRDYRFDMGFDYWGQGTTVTVSS',
    'H'
)
print(best)
# QVQLVQSGVEVKKPGASVKVSCKASGYTFTNYYMYWVRQAPGQGLEWMGGINPSNGGTNFNEKFKNRVTLTTDSSTTTAYMELKSLQFDDTAVYYCARRDYRFDMGFDYWGQGTTVTVSS

Suggesting mutations

Return residue scores for a given sequence:

import sapiens

scores = sapiens.predict_scores(
    '**QLV*SGVEVKKPGASVKVSCKASGYTFTNYYMYWVRQAPGQGLEWMGGINPSNGGTNFNEKFKNRVTLTTDSSTTTAYMELKSLQFDDTAVYYCARRDYRFDMGFDYWGQGTTVTVSS',
    'H'
)
scores.head()
#           A         C         D         E  ...
# 0  0.003272  0.004147  0.004011  0.004590  ... <- based on masked input
# 1  0.012038  0.003854  0.006803  0.008174  ... <- based on masked input
# 2  0.003384  0.003895  0.003726  0.004068  ... <- based on Q input
# 3  0.004612  0.005325  0.004443  0.004641  ... <- based on L input
# 4  0.005519  0.003664  0.003555  0.005269  ... <- based on V input
#
# Scores are given both for residues that are masked and that are present. 
# When inputting a non-human antibody sequence, the output scores can be used for humanization.

Antibody sequence embedding

Get a vector representation of each position in a sequence

import sapiens

residue_embed = sapiens.predict_residue_embedding(
    'QVKLQESGAELARPGASVKLSCKASGYTFTNYWMQWVKQRPGQGLDWIGAIYPGDGNTRYTHKFKGKATLTADKSSSTAYMQLSSLASEDSGVYYCARGEGNYAWFAYWGQGTTVTVSS', 
    'H', 
    layer=None
)
residue_embed.shape
# (layer, position in sequence, features)
# (5, 119, 128)

Get a single vector for each sequence

seq_embed = sapiens.predict_sequence_embedding(
    'QVKLQESGAELARPGASVKLSCKASGYTFTNYWMQWVKQRPGQGLDWIGAIYPGDGNTRYTHKFKGKATLTADKSSSTAYMQLSSLASEDSGVYYCARGEGNYAWFAYWGQGTTVTVSS', 
    'H', 
    layer=None
)
seq_embed.shape
# (layer, features)
# (5, 128)

Notebooks

Try out Sapiens in your browser using these example notebooks:

Links	Notebook	Description
	01_sapiens_antibody_infilling	Predict missing positions in an antibody sequence
	02_sapiens_antibody_embedding	Get vector representations and visualize them using t-SNE

Acknowledgements

Sapiens 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

Sapiens is a human antibody language model based on BERT.

Related tags

Overview

Sapiens: Human antibody language model

Features

Usage

Antibody sequence infilling

Suggesting mutations

Antibody sequence embedding

Notebooks

Acknowledgements

Owner

Merck Sharp & Dohme Corp. a subsidiary of Merck & Co., Inc.

A simple Streamlit App to classify swahili news into different categories.

This project uses unsupervised machine learning to identify correlations between daily inoculation rates in the USA and twitter sentiment in regards to COVID-19.

⚖️ A Statutory Article Retrieval Dataset in French.

Predicting the usefulness of reviews given the review text and metadata surrounding the reviews.

用Resnet101+GPT搭建一个玩王者荣耀的AI

Flaxformer: transformer architectures in JAX/Flax

ProteinBERT is a universal protein language model pretrained on ~106M proteins from the UniRef90 dataset.

Extract city and country mentions from Text like GeoText without regex, but FlashText, a Aho-Corasick implementation.

Write Alphabet, Words and Sentences with your eyes.

ADCS cert template modification and ACL enumeration

Easy Language Model Pretraining leveraging Huggingface's Transformers and Datasets

Tevatron is a simple and efficient toolkit for training and running dense retrievers with deep language models.

Global Rhythm Style Transfer Without Text Transcriptions

This project deals with a simplified version of a more general problem of Aspect Based Sentiment Analysis.

End-to-end MLOps pipeline of a BERT model for emotion classification.

Text to speech converter with GUI made in Python.

A full spaCy pipeline and models for scientific/biomedical documents.

YACLC - Yet Another Chinese Learner Corpus

An extension for asreview implements a version of the tf-idf feature extractor that saves the matrix and the vocabulary.

Python Implementation of ``Modeling the Influence of Verb Aspect on the Activation of Typical Event Locations with BERT'' (Findings of ACL: ACL 2021)