Protein Structure Tokenization via Geometric Byte Pair Encoding

Researchers have developed GeoBPE, a new protein structure tokenization method that converts protein backbone structures into discrete geometric tokens, achieving over 10x compression and data efficiency improvements. The approach uses geometry-grounded byte-pair encoding to create hierarchical vocabularies of protein structural primitives that align with functional families and enable better multimodal protein modeling.

• →GeoBPE achieves over 10x reduction in bits-per-residue compression while maintaining structural accuracy
• →The method requires 10x less training data compared to existing protein structure tokenization approaches
• →GeoBPE tokens align with CATH functional families, providing interpretable structural representations
• →The approach is architecture-agnostic and outperforms leading protein structure tokenizers across 12 tasks and 24 test splits
• →The method enables unconditional protein backbone generation through language modeling with transformers