Autoregressive language model generating de novo functional protein sequences
ProtGPT2 is a 738M parameter autoregressive transformer trained on 45 million protein sequences, generating de novo proteins that follow the principles of natural protein design.
Autoregressive unconditional and conditional sequence generationProduces sequences with natural secondary structure propensities and globular foldingTrained on UniRef50 with zero structural supervision requiredAvailable on Hugging Face Model Hub with straightforward pipeline inference
ProtGPT2 is a 738M parameter autoregressive transformer trained on 45 million protein sequences, generating de novo proteins that follow the principles of natural protein design.
Informazioni verificate rispetto a una fonte ufficiale; non un test pratico. Fonte · Ultima revisione: 20/09/2026, 11:10:15 UTC
Caratteristiche principali
Autoregressive unconditional and conditional sequence generation
Produces sequences with natural secondary structure propensities and globular folding
Trained on UniRef50 with zero structural supervision required
Available on Hugging Face Model Hub with straightforward pipeline inference
Interactive 3D Illustration
Biological Architecture (Stylized Ribbon)
Default view is a stylized demonstration ribbon. Enter any 4-letter RCSB PDB accession code above to stream real atomic coordinates.
⇄ Drag to rotate · Scroll to zoom
Illustrative Confidence Gradient:
>90 High / Low B 70-90 Confident 50-70 Low <50 Very low / High B
Academic Context & Research Evidence
Biological & Workflow Fit
Biological Application
De novo protein sequence generation and enzyme design
Research Workflow
Prompt with motif or sample unconditionally -> evaluate folding with AlphaFold/ESMFold
Use this citation format when referencing ProtGPT2 in scientific publications and benchmark papers.
@software{protgpt2_2026,
title = {{ProtGPT2}},
author = {{University of Bayreuth}},
year = {2026},
url = {https://huggingface.co/nferruz/ProtGPT2},
note = {Indexed on aibioatlas - AI for Biology and Drug Discovery}
}
Peer-Reviewed Literature & Preprints
Live scientific citations queried from Europe PMC and PubMed for ProtGPT2.
⏳ Querying Europe PMC & PubMed for indexed publications...
Informazioni tecniche / prodotto
Missing values mean the catalog has no recorded information. They do not mean a feature is absent.
Entry typeAI Model
Access modeOpen Source
AI roleGenerative Sequence Modeling
Input dataNot recorded
Output dataNot recorded
Licence conditionsApache 2.0
Commercial eligibilityOpen source release
Compute requirementsStandard GPU or multi-core CPU for inference
ValidationNot recorded
TypeAutoregressive protein transformer
Intended useNot recorded
CompatibilityNot recorded
ManufacturerUniversity of Bayreuth
Biological applicationDe novo protein sequence generation and enzyme design
Research workflowPrompt with motif or sample unconditionally -> evaluate folding with AlphaFold/ESMFold
Evidence levelPeer-reviewed research (Nature Communications 2022)