AbMAP Integration: Advanced Antibody Embeddings Now Available on Vecura
This update enables antibody researchers and protein engineers to generate highly specialized antibody embeddings through a streamlined workflow on Vecura, removing the need for complex, manual computational setups.
What is AbMAP?
AbMAP (Antibody Mutagenesis-Augmented Processing) is a specialized protein language model designed to overcome the limitations of general-purpose models when analyzing antibody sequences. By augmenting foundational protein language model embeddings with targeted in-silico mutagenesis of complementarity-determining regions (CDRs) and specific CDR-region isolation, it generates highly refined embeddings that better capture antibody-specific structural and functional signals. It is especially useful for researchers working on antibody structure prediction, binding affinity analysis, and B-cell repertoire profiling.
What can users do with AbMAP on Vecura?
With AbMAP on Vecura, users can:
- Convert heavy- or light-chain antibody variable-domain sequences into high-quality, dense vector embeddings.
- Select from multiple foundational protein language model backbones (such as ESM-2, ProtBert, or others) to customize the base embedding.
- Generate task-optimized representations by selecting between 'structure' or 'function' modes.
- Produce either fixed-length sequence vectors for global analysis or variable-length per-residue embeddings for position-sensitive research, without managing complex local infrastructure.
What the output means
The output provides a pickled dictionary of PyTorch tensors corresponding to the input sequences, alongside metadata including embedding shapes and process validation.
This output should be used to support scientific decision-making. It does not replace experimental validation.
Why this matters
General-purpose protein language models often struggle with the extreme sequence diversity found in antibody CDRs, which are critical for antigen binding. AbMAP bridges this gap by explicitly focusing the representation power of these models on the regions that matter most for antibody function.
By providing these advanced embeddings, AbMAP facilitates more accurate downstream machine learning tasks, such as predicting binding affinity or analyzing antibody diversity, ultimately accelerating the discovery and engineering of therapeutic antibodies.
- Developed by: Research group led by Singh and Im et al.
- Source: GitHub Repository and PNAS 2024 Publication
- Reference: AbMAP PNAS paper (Singh, Im et al. 2024)
Vecura で AbMAP を試す。
モデルワークスペースを開き、ご自身の入力で評価を始めましょう。