OPIG’s growing immunoinformatics team continues to develop and openly distribute a wide variety of databases and software packages for antibody/nanobody/T-cell receptor analysis. Below is a summary of all the latest updates (follows on from v1.0 and v2.0).

Databases

SAbDab (2014-present, incl. SAbDab-Nano, 2022-present): Our Structural Antibody Database continues to self-update and capture all the antibodies and nanobodies in the PDB (over 7,500 in total). All entries are cleaned, numbered, and annotated with metadata. STCRDab does the same for all the TCR structures in the PDB.
Paper 1 (SAbDab), Paper 2 (SAbDab-Nano), Paper 3 (STCRDab)

Observed Antibody Space (unpaired: 2018-present, paired: 2021-present): Our Observed Antibody Space database cleans and processes antibody repertoire sequencing datasets into an AIRR-compliant format, retaining useful patient metadata. It has recently been updated to capture paired sequence data. Currently contains c. 2.4Bn unpaired sequences and 1.5Mn paired sequences.
Paper 1, Paper 2

Thera-SAbDab (2019-present): Our Therapeutic Structural Antibody Databases contains every therapeutic antibody assigned a non-proprietary name by the World Health Organisation, with metadata and links to their structures in SAbDab. Updates twice-yearly. Currently contains over 850 entries.
Paper

CoV-AbDab (2020-present): Our Coronavirus Antibody Database tracks every antibody/nanobody in the literature with confirmed binding activity against a coronavirus. Variant/clade binding & neutralisation activities are recorded and updated over time. Currently contains over 12,500 entries.
Paper

PLAbDab (2023): New this week! Our Patent and Literature Antibody Database tracks down and pairs antibody sequences reported in papers/filings, with links to the original source literature. It can be mined by keyword, sequence, or structure. Currently contains over 65,000 sequence non-redundant entries — the largest set of functionally characterised antibody data.
Preprint


Tools

TAP (2019-present): Our Therapeutic Antibody Profiler tool is used for the computational developability assessment of candidate antibodies by comparing their 3D physicochemical properties with those of clinical-stage therapeutics. Guidelines evolve over time as more therapeutics reach Phase-II of clinical trials. Recently updated to use ABodyBuilder2 models.
Paper 1, Preprint 2

hu-mAb (2021): Our humanisation tool for antibodies, based on a random forest architecture trained on natural sequence data. Can be used to classify sequences as human or non-human, and has been used to suggest mutations to increase sequence humanness/decrease immunogenicity.
Paper

DLAB (2021): A pair of convolutional neural networks trained on voxelised representations of rigid-docked antibody model-antigen binding sites for improved high-throughput virtual screening.
Paper

Paragraph (2022): Our deep-learning based tool for paratope (antibody binding residue) prediction. Can be used to guide mutagenesis studies or for chemistry-based repertoire clustering within a paratyping framework.
Paper

AbLang (2022): A heavy and light chain language model for antibodies. An information-rich representation of antibody sequences that can improve predictive performance on downstream tasks.
Paper

ImmuneBuilder (2023): A suite of deep-learning based tools for accurately and quickly building 3D structural models of antibodies (ABodyBuilder2), nanobodies (NanoBodyBuilder2), and T-cell receptors (TCRBuilder2). They are now OPIG’s go-to tools for immunoglobulin structure prediction.
Paper

SPACE2 (2023, builds off SPACE from 2021): Our updated tool for antibody epitope profiling. Antigen-specific antibodies are structurally modelled (ABodyBuilder2) and clustered together (trained agglomerative clustering); those that are structurally similar are likely to bind to the same epitope.
Paper 1, Preprint 2

KA-Search (2023): Known Antibody-Search is a codebase for efficiently mining on the order of billions of sequences for antibodies that are similar to your query across any user-defined region/combination of residues (e.g. predicted paratope).
Preprint

Graphinity (2023): OPIG’s first foray into antibody-antigen affinity prediction. Graphinity has an equivariant graph neural network architecture built directly from antibody-antigen structures and achieves state-of-the-art performance on experimental ∆∆G prediction.
Preprint

—

A reminder that our webapp tools and databases are available under an academic or commercial licence via our SAbBox platform. This is now distributed as a Singularity Container (aca, com) as well as a vagrant virtual machine (aca, com).