AI Insight
Researchers developed MicroGlot, a DNA foundation model for microorganisms trained on 3.70 million sequences from 99,700 species totaling 378.3 billion nucleotides. The model incorporates taxonomic relationships through hyperbolic embeddings within a sparse mixture-of-experts architecture, enabling it to learn representations that capture both phenotypic traits and taxonomic identity. MicroGlot demonstrates improved performance over taxonomy-agnostic variants and achieves competitive results with lower computational costs by leveraging optimized training techniques from modern large language models.
Why it matters
This foundation model could accelerate microbial genomics research with applications in agriculture, biotechnology, and human health by providing more efficient and accurate analysis of microbial DNA sequences. The taxonomy-informed approach addresses the challenge of imbalanced representation across diverse microbial species, potentially improving predictions for understudied organisms.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Microorganisms are indispensable to terrestrial ecosystems, with their genomic material underpinning critical functions and applications across agriculture, biotechnology, and human health. Although genomic language models have advanced representation learning in DNA sequences, the extensive diversity of microorganisms and imbalanced taxonomic representation in the pretraining corpora pose challenges for effective microbial genomic sequence modeling. Here we present MicroGlot, a taxonomy-informed microbial DNA foundation model pretrained on 3.70 million sequences comprising 378.3 billion nucleotides across 99{,}700 species. MicroGlot encodes the hierarchical relations among taxa through hyperbolic embeddings, incorporating microbial taxonomic knowledge into a sparse mixture-of-experts architecture. Zero-shot evaluation of MicroGlot’s layer embeddings demonstrates that the model’s representations encode phenotypic traits and taxonomic identity. Comparison with a taxonomy-ablated variant trained under the same pretraining scheme shows that incorporating taxonomic knowledge consistently improves representation quality across the layers of MicroGlot. MicroGlot also combines optimized training techniques with efficient architectural components from modern large language models, achieving leading zero-shot performance across layers and competitive fine-tuning performance with low computational overhead. In a 1000-species set sampled from major cellular domains and viral realms, MicroGlot’s routing fingerprints show greater agreement with taxonomic groups than tetranucleotide composition, reflecting taxonomically structured expert routing in multilingual modeling of microbial genomes. Overall, we show that MicroGlot serves as an efficient and effective DNA foundation model for microbial genomic analysis.
Source: A Taxonomy-Informed Sparse DNA Foundation Model for Microbial Genomics