Efficient biological design via genomic in-context learning
Preprint
Abstract
In-context learning (ICL), the ability to perform new tasks from a few examples supplied in a prompt, is a defining emergent capability of large language models. Here, we show that ICL emerges in the genome language model Evo 2 and enables biological design. With as few as four examples in the prompt and no additional training, Evo 2 generates realistic metal-binding proteins, de novo antibodies, protein-RNA systems, and entire bacteriophage genomes. For ΦX174-like bacteriophage design, ICL surpasses previously fine-tuned models on computational measures of design quality and diversity. Beyond ΦX174, conditioning the model with different phage prompts enables us to efficiently design MS2 and M13 genomes in silico without fine-tuning. Iteratively updating the prompt with newly generated phage genomes, which we refer to as genomic chain-of-thought, further improves the diversity of designs while maintaining quality metrics. Paired with an inference engine that accelerates generation up to 35-fold and extends practicable context to a megabase, these results establish ICL as a versatile and efficient paradigm for biological sequence design.Keywords
Funding Information
- Stanford Human-Centered Artificial Intelligence
- Advanced Research Projects Agency for Health
- Schmidt Sciences AI2050
- Stanford Center for Digital Health
- Silicon Valley Community Foundation
- U.S. Department of Energy (Office of Biological and Environmental Research project award number SCW1929)
- U.S. Department of Energy (DE-AC52-07NA27344 (LLNL-JRNL-2026228))
- National Science Foundation
This publication has 51 references indexed in Scilit:
- Near-Adjacent Heavy Lanthanide Separation and Sensing Using Dimerizing LanmodulinsJournal of the American Chemical Society, 2026
- Diversity-driven biochemical survey reveals widespread dimerization throughout the rubisco superfamilyNature Communications, 2026
- Semantic design of functional de novo genes from a genomic language modelNature, 2025
- Accurate structure prediction of biomolecular interactions with AlphaFold 3Nature, 2024
- Enhanced rare-earth separation with a metal-sensitive lanmodulin dimerNature, 2023
- Lanmodulin: A Highly Selective Lanthanide-Binding Protein from a Lanthanide-Utilizing BacteriumJournal of the American Chemical Society, 2018
- MMseqs2 enables sensitive protein sequence searching for the analysis of massive data setsNature Biotechnology, 2017
- Multiobjective evolutionary algorithms to identify highly autocorrelated areas: the case of spatial distribution in financially compromised farmsAnnals of Operations Research, 2011
- Scoring function for automated assessment of protein structure template qualityProteins: Structure, Function, and Bioinformatics, 2004
- A general method applicable to the search for similarities in the amino acid sequence of two proteinsJournal of Molecular Biology, 1970