Peter Hyungjun Yoon
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Peter Hyungjun Yoon has a total of 101 co-authors. Here are the top co-authors by number of shared publications:
- Jennifer A. Doudna
Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, CA, USA
11 shared publications - Honglue Shi
Innovative Genomics Institute, University of California, Berkeley, Berkeley, CA, USA
7 shared publications - Kenneth J. Loi
Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, CA, USA
7 shared publications - Ron S. Boger
Innovative Genomics Institute, University of California, Berkeley, CA, USA
7 shared publications - Benjamin A. Adler
Innovative Genomics Institute, University of California, Berkeley, Berkeley, CA, USA
5 shared publications - Conner J. Langeberg
University of California, Berkeley
5 shared publications
Latest Publications
- Open Access
Prokaryotic homologs of the major vault protein form engineerable vault nanocapsules
Vaults are the largest ribonucleoprotein particles in eukaryotic cells, built from 78 copies of the major vault protein (MVP), yet their function remains unknown four decades after their discovery. MVP homologs occur in prokaryotes and bacteriophages, but whether they form vaults is unknown. Here, we show that prokaryotic and bacteriophage MVP homologs assemble into vault particles. We surveyed approximately 15 billion proteins, found MVP homologs in 29 bacterial and 4 archaeal phyla, and used size-exclusion chromatography to rapidly screen and isolate vault particles of eight homologs expressed in Escherichia coli . Cryo-electron microscopy structures of the phage homolog L05 at high resolution reveal a shell of 96 MVP chains enclosing a volume ∼25% larger than human vaults, among the largest protein shells of defined size outside virus capsids. Prokaryotic MVPs typically carry seven wall repeats and eukaryotic MVPs nine, and adding wall repeats to L05 yields larger vaults. Prokaryotic vaults expand the known diversity of vault architecture and offer a tractable system for studying vault assembly and a size-tunable scaffold for cargo encapsulation.
Open AccessProkaryotic homologs of the major vault protein form engineerable vault nanocapsules
Vaults are the largest ribonucleoprotein particles in eukaryotic cells, built from 78 copies of the major vault protein (MVP), yet their function remains unknown four decades after their discovery. MVP homologs occur in prokaryotes and bacteriophages, but whether they form vaults is unknown. Here, we show that prokaryotic and bacteriophage MVP homologs assemble into vault particles. We surveyed approximately 15 billion proteins, found MVP homologs in 29 bacterial and 4 archaeal phyla, and used size-exclusion chromatography to rapidly screen and isolate vault particles of eight homologs expressed in Escherichia coli . Cryo-electron microscopy structures of the phage homolog L05 at high resolution reveal a shell of 96 MVP chains enclosing a volume ∼25% larger than human vaults, among the largest protein shells of defined size outside virus capsids. Prokaryotic MVPs typically carry seven wall repeats and eukaryotic MVPs nine, and adding wall repeats to L05 yields larger vaults. Prokaryotic vaults expand the known diversity of vault architecture and offer a tractable system for studying vault assembly and a size-tunable scaffold for cargo encapsulation.
- Open Access
Autonomous AI agents discover reverse transcriptases with tandem repeat arrays
Metagenomic databases hold vast quantities of unexplored genetic material that can seed tools for biotechnology and medicine, but discovery is limited by our ability to examine sequence data at scale. Here, we show that large language models (LLMs) in an autonomous harness can alleviate this bottleneck and surface novel molecular systems. We deployed Claude Code instances to survey reverse transcriptase (RT) loci across 1.9 billion protein clusters. Among the top candidates were array-associated RTs (ART), a novel family of jumbo-phage RTs coupled with an array of ∼200-nucleotide units and a dedicated partner gene. Session transcripts show that Claude identified the ART family by directly examining its DNA sequences and recognizing the atypical repeats, a behavior attributable to specific Mythos 5 internal signals that respond to repeated DNA. ART arrays are highly expressed and appear as discrete units during Staphylococcus phage infection, suggesting an RT system directed by a repertoire of distinct RNAs. These findings indicate that LLMs can autonomously detect anomalies and drive analyses to initiate biological discoveries.
Open AccessAutonomous AI agents discover reverse transcriptases with tandem repeat arrays
Metagenomic databases hold vast quantities of unexplored genetic material that can seed tools for biotechnology and medicine, but discovery is limited by our ability to examine sequence data at scale. Here, we show that large language models (LLMs) in an autonomous harness can alleviate this bottleneck and surface novel molecular systems. We deployed Claude Code instances to survey reverse transcriptase (RT) loci across 1.9 billion protein clusters. Among the top candidates were array-associated RTs (ART), a novel family of jumbo-phage RTs coupled with an array of ∼200-nucleotide units and a dedicated partner gene. Session transcripts show that Claude identified the ART family by directly examining its DNA sequences and recognizing the atypical repeats, a behavior attributable to specific Mythos 5 internal signals that respond to repeated DNA. ART arrays are highly expressed and appear as discrete units during Staphylococcus phage infection, suggesting an RT system directed by a repertoire of distinct RNAs. These findings indicate that LLMs can autonomously detect anomalies and drive analyses to initiate biological discoveries.
- 17 September 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 393 (6817) , 1236-1240
VIPR RNA-guided DNA recognition by noncontiguous geometric triplex formation
Viral interference programmable repeat (VIPR) systems use a noncontiguous code for RNA-guided transcriptional silencing. How the Vipr protein and a VIPR RNA (vrRNA) comprising alternating GGY and NN segments achieve precise DNA targeting is unknown. Here, we present 21 cryo–electron microscopy structures that help explain the mechanism of target engagement. Vipr protomers oligomerize along the vrRNA to form a right-handed helical filament, sequestering each GGY motif and positioning the adjacent NN bases for target base pairing. DNA binding, in which every third nucleotide is skipped, results in a gapped vrRNA-DNA hybrid helix that encircles the nontarget DNA strand to form a geometric triplex. These findings suggest that triplex-mediated target-strand handoff could enable noncontiguous and programmable RNA-guided DNA recognition in VIPR systems.
- 17 September 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 393 (6817) , 1236-1240
VIPR RNA-guided DNA recognition by noncontiguous geometric triplex formation
Viral interference programmable repeat (VIPR) systems use a noncontiguous code for RNA-guided transcriptional silencing. How the Vipr protein and a VIPR RNA (vrRNA) comprising alternating GGY and NN segments achieve precise DNA targeting is unknown. Here, we present 21 cryo–electron microscopy structures that help explain the mechanism of target engagement. Vipr protomers oligomerize along the vrRNA to form a right-handed helical filament, sequestering each GGY motif and positioning the adjacent NN bases for target base pairing. DNA binding, in which every third nucleotide is skipped, results in a gapped vrRNA-DNA hybrid helix that encircles the nontarget DNA strand to form a geometric triplex. These findings suggest that triplex-mediated target-strand handoff could enable noncontiguous and programmable RNA-guided DNA recognition in VIPR systems.
- 17 September 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 393 (6817) , 1230-1235
A noncontiguous code for RNA-guided DNA recognition at the origin of CRISPR-Cas
CRISPR-Cas provides RNA-mediated adaptive immunity, but how its first RNA-guided effector arose is unclear. In this study, we report the discovery of Viral Interference Programmable Repeat (VIPR) systems consisting of a Vipr protein ancestral to the earliest CRISPR-Cas effectors and VIPR RNAs (vrRNAs) comprising alternating GGY/NN motifs. Unlike canonical guide RNAs that pair with target nucleic acids through contiguous complementarity, vrRNAs recognize double-stranded DNA through a noncontiguous code in which the variable NN dinucleotides collectively specify a gapped target sequence. Natural vrRNA targets suggest that VIPR systems act against competing phages, and we demonstrate programmable phage defense by redirecting the complex for transcriptional repression. These results suggest that adaptive immunity originated from ancient warfare between viruses, revealing a previously unidentified logic for encoding information in sequence.
- 17 September 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 393 (6817) , 1230-1235
A noncontiguous code for RNA-guided DNA recognition at the origin of CRISPR-Cas
CRISPR-Cas provides RNA-mediated adaptive immunity, but how its first RNA-guided effector arose is unclear. In this study, we report the discovery of Viral Interference Programmable Repeat (VIPR) systems consisting of a Vipr protein ancestral to the earliest CRISPR-Cas effectors and VIPR RNAs (vrRNAs) comprising alternating GGY/NN motifs. Unlike canonical guide RNAs that pair with target nucleic acids through contiguous complementarity, vrRNAs recognize double-stranded DNA through a noncontiguous code in which the variable NN dinucleotides collectively specify a gapped target sequence. Natural vrRNA targets suggest that VIPR systems act against competing phages, and we demonstrate programmable phage defense by redirecting the complex for transcriptional repression. These results suggest that adaptive immunity originated from ancient warfare between viruses, revealing a previously unidentified logic for encoding information in sequence.
- 16 July 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 393 (6808) , 313-318
Structure and evolution-guided design of minimal RNA-guided nucleases
The design of RNA-guided nucleases with properties not limited by evolution can expand programmable genome-editing capabilities. However, generating diverse multidomain proteins with robust enzymatic properties remains challenging. Here, we use a protein design strategy that couples a structure-guided inverse-folding model with evolution-informed residue constraints to generate active, divergent variants of TnpB, a minimal CRISPR-Cas12–like nuclease, termed SynTnpBs. High-throughput screening of artificial intelligence–generated variants yielded editors that retained or exceeded wild-type activity in bacterial, plant, and human cells. Cryo–electron microscopy–based structure determination of the most divergent variant revealed stabilizing contacts in the RNA–DNA interfaces across conformations, demonstrating the design potential of this approach. Together, these results establish a strategy for creating non-natural RNA-guided nucleases and conformationally active nucleic acid binders, enlarging the designable protein space.
- 16 July 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 393 (6808) , 313-318
Structure and evolution-guided design of minimal RNA-guided nucleases
The design of RNA-guided nucleases with properties not limited by evolution can expand programmable genome-editing capabilities. However, generating diverse multidomain proteins with robust enzymatic properties remains challenging. Here, we use a protein design strategy that couples a structure-guided inverse-folding model with evolution-informed residue constraints to generate active, divergent variants of TnpB, a minimal CRISPR-Cas12–like nuclease, termed SynTnpBs. High-throughput screening of artificial intelligence–generated variants yielded editors that retained or exceeded wild-type activity in bacterial, plant, and human cells. Cryo–electron microscopy–based structure determination of the most divergent variant revealed stabilizing contacts in the RNA–DNA interfaces across conformations, demonstrating the design potential of this approach. Together, these results establish a strategy for creating non-natural RNA-guided nucleases and conformationally active nucleic acid binders, enlarging the designable protein space.
- Open Access
- 8 January 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 391 (6781) , 195-201
Recurrent acquisition of nuclease-protease pairs in antiviral immunity
Antiviral immune systems diversify by integrating new genes into existing pathways, creating new mechanisms of viral resistance. We identified genes encoding a predicted nuclease paired with a trypsin-like protease repeatedly acquired by multiple, otherwise unrelated antiviral immune systems in bacteria. Cell-based and biochemical assays revealed that the nuclease is a proenzyme that cleaves DNA only after activation by its partner protease. Two distinct immune systems, Hachiman and AVAST (antiviral adenosine triphosphatase/nucleoside triphosphatase of the STAND superfamily, Avs), use the same mechanism of proteolytic activation despite their independent evolutionary origins. Examination of nuclease-protease inheritance patterns identified caspase-nuclease ( canu ) genomic loci that confer antiviral defense in a pathway reminiscent of eukaryotic caspase activation. These results uncover the coordinated activities of pronucleases and their activating proteases within different immune systems and show how coevolution enables defense system innovation.
- 8 January 2026
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 391 (6781) , 195-201
Open AccessRecurrent acquisition of nuclease-protease pairs in antiviral immunity
Antiviral immune systems diversify by integrating new genes into existing pathways, creating new mechanisms of viral resistance. We identified genes encoding a predicted nuclease paired with a trypsin-like protease repeatedly acquired by multiple, otherwise unrelated antiviral immune systems in bacteria. Cell-based and biochemical assays revealed that the nuclease is a proenzyme that cleaves DNA only after activation by its partner protease. Two distinct immune systems, Hachiman and AVAST (antiviral adenosine triphosphatase/nucleoside triphosphatase of the STAND superfamily, Avs), use the same mechanism of proteolytic activation despite their independent evolutionary origins. Examination of nuclease-protease inheritance patterns identified caspase-nuclease ( canu ) genomic loci that confer antiviral defense in a pathway reminiscent of eukaryotic caspase activation. These results uncover the coordinated activities of pronucleases and their activating proteases within different immune systems and show how coevolution enables defense system innovation.
- Open Access
Structure and evolution-guided design of minimal RNA-guided nucleases
The design of RNA-guided nucleases with properties not limited by evolution can expand programmable genome editing capabilities. However, generating diverse multi-domain proteins with robust enzymatic properties remains challenging. Here we use an artificial intelligence-driven strategy that couples structure-guided inverse protein folding with evolution-informed residue constraints to generate active, divergent variants of TnpB, a minimal CRISPR-Cas12-like nuclease. High-throughput functional screening of AI-generated variants yielded editors that retained or exceeded wild-type activity in bacterial, plant and human cells. Cryo-EM-based structure determination of the most divergent active variant revealed new stabilizing contacts in the RNA/DNA interfaces across conformational states, demonstrating the design potential of this approach. Together these results establish a strategy for creating non-natural RNA-guided nucleases and conformationally active nucleic acid binders, enlarging the designable protein space.
Open AccessStructure and evolution-guided design of minimal RNA-guided nucleases
The design of RNA-guided nucleases with properties not limited by evolution can expand programmable genome editing capabilities. However, generating diverse multi-domain proteins with robust enzymatic properties remains challenging. Here we use an artificial intelligence-driven strategy that couples structure-guided inverse protein folding with evolution-informed residue constraints to generate active, divergent variants of TnpB, a minimal CRISPR-Cas12-like nuclease. High-throughput functional screening of AI-generated variants yielded editors that retained or exceeded wild-type activity in bacterial, plant and human cells. Cryo-EM-based structure determination of the most divergent active variant revealed new stabilizing contacts in the RNA/DNA interfaces across conformational states, demonstrating the design potential of this approach. Together these results establish a strategy for creating non-natural RNA-guided nucleases and conformationally active nucleic acid binders, enlarging the designable protein space.
- Open Access
A miniature CRISPR-Cas10 enzyme confers immunity by an inverse signaling pathway
Microbial and viral co-evolution has created immunity mechanisms involving oligonucleotide signaling that share mechanistic features with human anti-viral systems 1 . In these pathways, including CBASS and type III CRISPR systems in bacteria and cGAS-STING in humans, oligonucleotide synthesis occurs upon detection of virus or foreign genetic material in the cell, triggering the antiviral response 2–4 . In a surprising inversion of this process, we show here that the CRISPR-related enzyme mCpol synthesizes cyclic oligonucleotides constitutively as part of an active mechanism that maintains cell health. Cell-based experiments demonstrated that the absence or loss of mCpol-produced cyclic oligonucleotides triggers cell death, preventing spread of viruses that attempt immune evasion by depleting host cyclic nucleotides. Structural and mechanistic investigation revealed mCpol to be a di-adenylate cyclase whose product, c-di-AMP, prevents toxic oligomerization of the effector protein 2TMβ. Analysis of cells by fluorescence microscopy showed that lack of mCpol allows 2TMβ-mediated cell death due to inner membrane collapse. These findings unveil a powerful new defense strategy against virus-mediated immune suppression, expanding our understanding of oligonucleotides in cell health and disease. These results raise the possibility of similar protective roles for cyclic oligonucleotides in other organisms including humans.
Open AccessA miniature CRISPR-Cas10 enzyme confers immunity by an inverse signaling pathway
Microbial and viral co-evolution has created immunity mechanisms involving oligonucleotide signaling that share mechanistic features with human anti-viral systems 1 . In these pathways, including CBASS and type III CRISPR systems in bacteria and cGAS-STING in humans, oligonucleotide synthesis occurs upon detection of virus or foreign genetic material in the cell, triggering the antiviral response 2–4 . In a surprising inversion of this process, we show here that the CRISPR-related enzyme mCpol synthesizes cyclic oligonucleotides constitutively as part of an active mechanism that maintains cell health. Cell-based experiments demonstrated that the absence or loss of mCpol-produced cyclic oligonucleotides triggers cell death, preventing spread of viruses that attempt immune evasion by depleting host cyclic nucleotides. Structural and mechanistic investigation revealed mCpol to be a di-adenylate cyclase whose product, c-di-AMP, prevents toxic oligomerization of the effector protein 2TMβ. Analysis of cells by fluorescence microscopy showed that lack of mCpol allows 2TMβ-mediated cell death due to inner membrane collapse. These findings unveil a powerful new defense strategy against virus-mediated immune suppression, expanding our understanding of oligonucleotides in cell health and disease. These results raise the possibility of similar protective roles for cyclic oligonucleotides in other organisms including humans.
- Open Access
- 5 December 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-17
RNA language models predict mutations that improve RNA function
Structured RNA lies at the heart of many central biological processes, from gene expression to catalysis. RNA structure prediction is not yet possible due to a lack of high-quality reference data associated with organismal phenotypes that could inform RNA function. We present GARNET (Gtdb Acquired RNa with Environmental Temperatures), a new database for RNA structural and functional analysis anchored to the Genome Taxonomy Database (GTDB). GARNET links RNA sequences to experimental and predicted optimal growth temperatures of GTDB reference organisms. Using GARNET, we develop sequence- and structure-aware RNA generative models, with overlapping triplet tokenization providing optimal encoding for a GPT-like model. Leveraging hyperthermophilic RNAs in GARNET and these RNA generative models, we identify mutations in ribosomal RNA that confer increased thermostability to the Escherichia coli ribosome. The GTDB-derived data and deep learning models presented here provide a foundation for understanding the connections between RNA sequence, structure, and function.
- 5 December 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-17
Open AccessRNA language models predict mutations that improve RNA function
Structured RNA lies at the heart of many central biological processes, from gene expression to catalysis. RNA structure prediction is not yet possible due to a lack of high-quality reference data associated with organismal phenotypes that could inform RNA function. We present GARNET (Gtdb Acquired RNa with Environmental Temperatures), a new database for RNA structural and functional analysis anchored to the Genome Taxonomy Database (GTDB). GARNET links RNA sequences to experimental and predicted optimal growth temperatures of GTDB reference organisms. Using GARNET, we develop sequence- and structure-aware RNA generative models, with overlapping triplet tokenization providing optimal encoding for a GPT-like model. Leveraging hyperthermophilic RNAs in GARNET and these RNA generative models, we identify mutations in ribosomal RNA that confer increased thermostability to the Escherichia coli ribosome. The GTDB-derived data and deep learning models presented here provide a foundation for understanding the connections between RNA sequence, structure, and function.
- Open Access
- 2 August 2024
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 385 (6708) , 538-543
Structure-guided discovery of ancestral CRISPR-Cas13 ribonucleases
The RNA-guided ribonuclease CRISPR-Cas13 enables adaptive immunity in bacteria and programmable RNA manipulation in heterologous systems. Cas13s share limited sequence similarity, hindering discovery of related or ancestral systems. To address this, we developed an automated structural-search pipeline to identify an ancestral clade of Cas13 (Cas13an) and further trace Cas13 origins to defense-associated ribonucleases. Despite being one-third the size of other Cas13s, Cas13an mediates robust programmable RNA depletion and defense against diverse bacteriophages. However, unlike its larger counterparts, Cas13an uses a single active site for both CRISPR RNA processing and RNA-guided cleavage, revealing that the ancestral nuclease domain has two modes of activity. Discovery of Cas13an deepens our understanding of CRISPR-Cas evolution and expands opportunities for precision RNA editing, showcasing the promise of structure-guided genome mining.
- 2 August 2024
- Published by American Association for the Advancement of Science (AAAS) in Science
- Vol. 385 (6708) , 538-543
Open AccessStructure-guided discovery of ancestral CRISPR-Cas13 ribonucleases
The RNA-guided ribonuclease CRISPR-Cas13 enables adaptive immunity in bacteria and programmable RNA manipulation in heterologous systems. Cas13s share limited sequence similarity, hindering discovery of related or ancestral systems. To address this, we developed an automated structural-search pipeline to identify an ancestral clade of Cas13 (Cas13an) and further trace Cas13 origins to defense-associated ribonucleases. Despite being one-third the size of other Cas13s, Cas13an mediates robust programmable RNA depletion and defense against diverse bacteriophages. However, unlike its larger counterparts, Cas13an uses a single active site for both CRISPR RNA processing and RNA-guided cleavage, revealing that the ancestral nuclease domain has two modes of activity. Discovery of Cas13an deepens our understanding of CRISPR-Cas evolution and expands opportunities for precision RNA editing, showcasing the promise of structure-guided genome mining.