About COMPASS

The stretches of DNA between protein-coding genes are often called the genome’s “dark matter,” but they are far from empty. They encode RNA molecules that sense metabolites, regulate genes, and guide programmable genome editors such as CRISPR and bridge RNAs. Yet much of this biology remains unexplored. One reason is that RNAs can retain similar structures and functions even as their DNA sequences change dramatically, making their relationships difficult to recognize with conventional computational tools.

To help map this territory, we developed Evo-PIN (Prokaryotic INtergenic), a version of the genomic language model Evo 2 fine-tuned on bacterial and archaeal intergenic sequences. Evo-PIN recognizes relationships between these regions even when they share little sequence similarity. We used it to organize 376 million intergenic sequences from more than 140,000 microbial genomes and large collections of metagenomic data into COMPASS (COMPrehensive Atlas of Structured Sequences), our atlas of noncoding sequences. By grouping related sequences and examining their shared RNA structures and neighboring genes, COMPASS provides a starting point for the discovery of new molecular functions.

This website features more than 250,000 selected clusters with candidate structured RNAs and genomic context assigned to each of them. Exploring a resource this large also calls for new ways of doing research. Alongside the website, we provide a Model Context Protocol (MCP) server that lets AI agents search the full database of over 5 million clusters and investigate candidates together with a researcher or fully autonomously. For more details on our methods and initial findings, refer to our preprint on bioRxiv.

@article{sun2026compass,
  title = {Systematic mapping of the prokaryotic noncoding sequence space with a genomic language model},
  author = {Sun, Gwanggyu and Hudelmaier, Tim and Gilligan, Joshua C. and Gonzalez, Gabriel A. and Yin, Peter and Wan, Fangping and Ma, Zijian (Carl) and Durrant, Matthew G. and Ricci-Tam, Chiara and Pawluk, April and Puschnik, Andreas S. and Konermann, Silvana and Hsu, Patrick D.},
  journal = {bioRxiv},
  year = {2026},
  doi = {10.64898/2026.10.21.760802},
  url = {https://compass.arcinstitute.org}
}