Accelerate Antibody Discovery with AbGPT on Vecura
This update enables immunologists and biotherapeutics researchers to perform de novo antibody design through a guided, high-throughput workflow inside Vecura, eliminating the need for complex local infrastructure setup.
What is AbGPT?
AbGPT is a generative language model based on the GPT-2 architecture, specifically fine-tuned on curated B-cell receptor (BCR) sequence data. It performs de novo design of antibody heavy and light chain amino acid sequences by autoregressively sampling from short, user-provided residue prompts.
It helps users rapidly generate large, diverse antibody libraries for immunological research without relying on traditional template-based design or experimental screening. It is especially useful for researchers looking to explore antibody sequence space efficiently in in-silico environments.
What can users do with AbGPT on Vecura?
With AbGPT on Vecura, users can:
- Generate novel, full-length BCR heavy or light chain sequences starting from simple 4-residue amino acid prompts.
- Create large, exploratory antibody panels or targeted variants around known germline seeds.
- Customize generation parameters like top-k sampling and repetition penalties to balance sequence diversity and validity.
- Streamline their research pipeline by using pre-warmed inference, avoiding the overhead of manual model management.
What the output means
The output provides a structured table of generated BCR amino-acid sequences, including the original prompt, the resulting sequence, and its length.
This output should be used to support scientific decision making. It does not replace experimental validation.
Why this matters
The design of novel antibodies is a cornerstone of modern biotherapeutics and immunology. Traditional methods for discovering and engineering antibodies are often time-consuming, expensive, and limited by template availability or existing experimental datasets.
AbGPT provides a powerful, automated approach to explore the vast space of potential BCR sequences. By enabling de novo generation, researchers can access a broader diversity of potential binding sequences, facilitating faster candidate discovery and deeper insights into antibody structure and repertoire development.
- Developed by: deskk
- Source: GitHub
- Reference: bioRxiv preprint
Try AbGPT on Vecura.
Open the model workspace and start evaluating it with your own inputs.