Systematic mapping of the prokaryotic noncoding sequence space with a genomic language model
- 1Arc Institute; Palo Alto, CA, USA
Abstract
Prokaryotic genomes encode a vast repertoire of noncoding elements such as ribozymes, riboswitches, and CRISPR RNAs. The sequence-level diversity of these intergenic elements precludes alignment-based clustering that would enable annotation and discovery of novel noncoding RNA families. Here, we show that a genomic language model fine-tuned on prokaryotic intergenic sequences learns representations of noncoding elements from DNA sequence. We used these embeddings to organize 376 million intergenic regions from 1.3 terabase pairs of prokaryotic genomes and metagenomes into 5.14 million clusters. This approach nominated 20,149 clusters of putative structured RNAs with a candidate protein partner. Periodicity in token entropies revealed internal architecture, recovering CRISPR, TIGR and VIPR arrays together with their cognate effector proteins from sequence alone. Targeted clustering of bridge recombinase-adjacent sequences identified bridge RNA variants with longer guide recognition sequences than any characterized ortholog. Finally, we directed teams of AI agents to mine the database under researcher-defined criteria, uncovering divergent riboswitches, ribozymes, and other structured RNAs with biological hypotheses for their function. We release a publicly available database that transforms the dark matter of prokaryotic genomes into an explorable resource for human-agent molecular discovery.
Related
- Evaluating the added value of sequence models in noncoding RNA assays: a baseline-aware framework
- DeepCDS: Ab initio coding sequence prediction in prokaryotic short nucleotide sequences
- Coevolutionary mining of prokaryotic non-coding elements with a genome language model
- Free‐breathing multislice native myocardial T 1 mapping using the slice‐interleaved T 1 (STONE) sequence
- From neuropeptide and receptor annotation to ligand-receptor pairing: a sequence- and structure-based framework for mapping the neuropeptide-receptor interactome in Gryllus bimaculatus
BibTeX
@misc{sun2026systematic,
title = {{Systematic mapping of the prokaryotic noncoding sequence space with a genomic language model}},
author = {G. Sun and T. Hudelmaier and J. C. Gilligan and G. A. Gonzalez and P. Yin and F. Wan and Z. Ma and M. G. Durrant and C. Ricci-Tam and A. Pawluk and A. S. Puschnik and S. Konermann and P. D. Hsu},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.21.760802},
url = {https://www.biorxiv.org/content/10.64898/2026.10.21.760802v1},
note = {CC BY-ND},
}
RIS
TY - UNPB
TI - Systematic mapping of the prokaryotic noncoding sequence space with a genomic language model
AU - G. Sun
AU - T. Hudelmaier
AU - J. C. Gilligan
AU - G. A. Gonzalez
AU - P. Yin
AU - F. Wan
AU - Z. Ma
AU - M. G. Durrant
AU - C. Ricci-Tam
AU - A. Pawluk
AU - A. S. Puschnik
AU - S. Konermann
AU - P. D. Hsu
PY - 2026
DA - 2026/10/21
PB - bioRxiv
DO - 10.64898/2026.10.21.760802
UR - https://www.biorxiv.org/content/10.64898/2026.10.21.760802v1
AB - Prokaryotic genomes encode a vast repertoire of noncoding elements such as ribozymes, riboswitches, and CRISPR RNAs. The sequence-level diversity of these intergenic elements precludes alignment-based clustering that would enable annotation and discovery of novel noncoding RNA families. Here, we show that a genomic language model fine-tuned on prokaryotic intergenic sequences learns representations of noncoding elements from DNA sequence. We used these embeddings to organize 376 million intergenic regions from 1.3 terabase pairs of prokaryotic genomes and metagenomes into 5.14 million clusters. This approach nominated 20,149 clusters of putative structured RNAs with a candidate protein partner. Periodicity in token entropies revealed internal architecture, recovering CRISPR, TIGR and VIPR arrays together with their cognate effector proteins from sequence alone. Targeted clustering of bridge recombinase-adjacent sequences identified bridge RNA variants with longer guide recognition sequences than any characterized ortholog. Finally, we directed teams of AI agents to mine the database under researcher-defined criteria, uncovering divergent riboswitches, ribozymes, and other structured RNAs with biological hypotheses for their function. We release a publicly available database that transforms the dark matter of prokaryotic genomes into an explorable resource for human-agent molecular discovery.
ER -