Literature · opig
p-IgGen
An autoregressive protein language model for designing diverse paired antibody heavy and light chain sequences.
What p-IgGen is
p-IgGen is an autoregressive protein language model developed by the Oxford Protein Informatics Group (OPIG) and Ollie Turnbull, with funding support from EPSRC and AstraZeneca. The model is specifically created for antibody sequence design and paired antibody library generation.
Designing therapeutic antibodies often requires coordinating the sequences of both the heavy and light chains that make up an immunoglobulin. By generating diverse paired antibody sequences, p-IgGen supports computational drug discovery and protein design workflows.
How it works
p-IgGen uses a GPT-NeoX autoregressive language model architecture adapted to handle protein sequences. It operates by generating sequence completions from text prompts.
The model was initially pre-trained on unpaired immunoglobulin heavy and light chain sequences sourced from the Observed Antibody Space (OAS) database. It was subsequently fine-tuned on paired antibody sequences to learn joint sequence patterns across both chains.
Key facts
- Architecture
- GPT-NeoX
- Pre-training data
- Unpaired heavy and light chains from the Observed Antibody Space (OAS)
- Fine-tuning data
- Paired antibody sequences
- Modality
- Protein sequence (text)
- Hardware requirement
- 16 GB GPU memory
What you can use it for
- Paired Antibody Library Generation
- Generating diverse synthetic libraries of paired heavy and light immunoglobulin chains for antibody discovery.
- Antibody Sequence Completion
- Producing candidate antibody sequences from prompt sequences via autoregressive sequence generation.
- Developability Modelling
- Generating antibody sequences and representations for downstream antibody developability evaluation pipelines.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
p-IgGen
Strengths and limits
Good at
- Generates paired heavy and light immunoglobulin chains rather than isolated single chains.
- Leverages comprehensive repertoire data from the Observed Antibody Space for pre-training.
- Built upon the established GPT-NeoX autoregressive framework.
Watch out for
- Requires 16 GB of GPU memory for serving and execution.
- Operates strictly at the primary sequence level without incorporating explicit structural coordinate data.
Inputs and outputs
You provide
Prompt
The model continues from here.
You get back
The model's completion.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use p-IgGen commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
How do I run p-IgGen without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on CPU only, typically in 12s, and a run costs about 28 credits. Visitors without an account get 1,000 credits a month.
What sequence data was used to train p-IgGen?
The model was pre-trained on unpaired heavy and light chain antibody sequences from the Observed Antibody Space (OAS) and then fine-tuned on paired sequences.
What architecture does p-IgGen use?
It is an autoregressive generative model built using the GPT-NeoX architecture.
What are the model inputs and outputs?
The model takes a text prompt representing an antibody sequence fragment and outputs a generated text completion.
How to cite it
academic.oup.com/bioinformatics/article/40/11/btae659/7888884
Written from p-IgGen's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.