It is shown that a genomic language model fine-tuned on prokaryotic intergenic sequences learns representations of noncoding elements from DNA sequence, which transforms the dark matter of prokaryotic genomes into an explorable resource for human-agent molecular discovery.
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.
Microbial genomes encode non-coding RNAs (ncRNAs) and nucleic-acid-interacting proteins essential to biology and biotechnology. However, discovery of these systems remains protein-centric and based on similarity to known sequences. Here, we introduce Minerva, a framework for coevolutionary mining that uses genome langu...
David B. Li, G. Brixi, Alexandra S. Kim et al.· bioRxiv· 4 citations
A computational study to detect functional RNA structures within the ENCODE regions of the human genome using three recently introduced programs based on either phylogenetic–stochastic context-free grammar (EvoFold, RNAz and AlifoldZ) or energy directed folding (RNAz and AlifoldZ), yielding several thousand candidate s...
R. Guigó, M. Snyder, M. Gerstein et al.· 0 citations
Many functional RNAs were first found as conserved RNA structures, and their roles in gene regulation were identified later. In humans and other vertebrates, imperfect genome alignments have hindered the discovery of such structures. Here we present RLMHelix, which selects candidate conserved RNA structures from single...
Shunsuke Sumi, Koki Makino, Masahiro Onoguchi et al.· bioRxiv· 0 citations
Non-coding RNAs (ncRNAs) regulate gene expression through diverse structural and sequence-based mechanisms, often acting without protein intermediates. Riboswitches are cis-regulatory ncRNAs that directly sense small molecule metabolites to control gene expression. Many riboswitches operate through kinetic control, mea...