EvoLen: Evolution-Guided Tokenization for DNA Language Model
2026-04-09 • Machine Learning
Machine Learning
AI summaryⓘ
The authors study how to break down DNA into meaningful pieces called tokens for computer models. They show that unlike human language, DNA doesn't have clear token boundaries, so choosing these tokens is important. Their method, EvoLen, uses evolutionary information to find and keep important biological patterns across species when creating tokens. Experiments show their approach better captures functional DNA parts and performs well on DNA modeling tasks compared to usual methods. This suggests adding evolution-based knowledge helps make DNA models more biologically relevant.
DNA tokenizationDNA language modelsbyte-pair encoding (BPE)regulatory motifsevolutionary constraintcross-species comparisonfunctional sequence patternsinductive biasdynamic programmingvocabulary merging
Authors
Nan Huang, Xiaoxiao Zhou, Junxia Cui, Mario Tapia-Pacheco, Tiffany Amariuta, Yang Li, Jingbo Shang
Abstract
Tokens serve as the basic units of representation in DNA language models (DNALMs), yet their design remains underexplored. Unlike natural language, DNA lacks inherent token boundaries or predefined compositional rules, making tokenization a fundamental modeling decision rather than a naturally specified one. While existing approaches like byte-pair encoding (BPE) excel at capturing token structures that reflect human-generated linguistic regularities, DNA is organized by biological function and evolutionary constraint rather than linguistic convention. We argue that DNA tokenization should prioritize functional sequence patterns like regulatory motifs-short, recurring segments under evolutionary constraint and typically preserved across species. We incorporate evolutionary information directly into the tokenization process through EvoLen, a tokenizer that combines evolutionary stratification with length-aware decoding to better preserve motif-scale functional sequence units. EvoLen uses cross-species evolutionary signals to group DNA sequences, trains separate BPE tokenizers on each group, merges the resulting vocabularies via a rule prioritizing preserved patterns, and applies length-aware decoding with dynamic programming. Through controlled experiments, EvoLen improves the preservation of functional sequence patterns, differentiation across genomic contexts, and alignment with evolutionary constraint, while matching or outperforming standard BPE across diverse DNALM benchmarks. These results demonstrate that tokenization introduces a critical inductive bias and that incorporating evolutionary information yields more biologically meaningful and interpretable sequence representations.