Seth L. Shipman

Advances in scientific AI have accelerated hypothesis generation, prediction, and experimental design, but translating these capabilities into reliable physical experimentation remains a major bottleneck in scientific discovery. Although automation, high-throughput experimentation, foundation models, and cloud infrastructure have accelerated individual stages of the scientific workflow, they have not unified the discovery process. Scientists must still coordinate between computational models and physical laboratory systems to translate scientific designs into executable experiments and ensure their results are reliable. We argue that the next generation of laboratories will be agentic: environments in which scientists, AI systems, equipment and robotic platforms operate as collaborative discovery partners. We propose an architecture built around two complementary components: an agentic foundation that translates scientific designs into coordinated execution plans, and an evolving laboratory world model that represents physical state, predicts action outcomes, and incorporates experimental feedback. Together, they connect computational reasoning to observable and verifiable physical execution. Multimodal perception, robotics, and human–AI interfaces can make laboratory actions and evolving experimental conditions accessible to these systems, supporting adaptive execution and learning. We further propose an evaluation framework centered on physical grounding, execution fidelity, scientific validity, and adaptive coordination under human oversight, rather than autonomy alone. This Perspective outlines how agentic laboratories could connect scientific reasoning, physical experimentation, and trustworthy evidence within a human-in-the-lead discovery loop.

Programmable RNA sensors enable sequence-specific detection of RNA inputs for diagnostics, genetic circuit control, and multiplexed molecular readouts. Although readily programmable, their performance varies substantially with the target site selected within a transcript, making target-site selection a major challenge. Here, we report DeSTAR (Detectron Sensor TARget-site ranker), a machine-learning-assisted target-site ranking model for programmable RNA sensors. To train this model, we generated transcript-wide, single-nucleotide-resolution tiling sensor libraries across six diverse RNA targets, enabling generalization to be tested across transcripts rather than within a single RNA. In leave-one-transcript-out evaluation, DeSTAR's top-ranked sites were enriched for high-performing sensors in held-out transcripts, driven predominantly by interpretable sequence grammar in the sensor stem rather than transcript-specific patterns. Prospectively, on an unseen SARS-CoV-2 transcript, the model's top-ranked sites yielded working sensors roughly five-fold more often than random selection. These results support DeSTAR as a generalizable strategy for target-site selection in RNA sensors across transcripts.

Programmable RNA sensors such as toehold switches are used to detect specific RNA sequences. However, their reliance on protein-based or RNA-based outputs limits their use in multiplexed and sequencing-based applications. Here we introduce Detectrons, modular biosensors that couple programmable toehold switches with retron-mediated reverse transcription to transduce RNA inputs into unique DNA barcodes, converting dynamic RNA signals into durable DNA records within living cells. The framework enables alternative modes of transcript-based sensing with applications including viral infection detection. Through the construction of a synthetic toehold retron library and application of machine learning, we uncover key design principles that improve signal strength and specificity. We apply Detectrons to the multiplexed live-cell detection of specific phage infections, enabling transcript-triggered barcode synthesis and quantitative host susceptibility profiling in pooled bacterial populations. Detectrons provide a scalable and generalizable strategy for phage screening and for recording transcriptional events in complex bacterial communities.

Bacterial defense systems provide a rich reservoir for biotechnological innovation. Retrons are tripartite abortive infection systems that detect phage invasion using reverse-transcribed DNA (msDNA), but how they structurally couple threat detection to effector activation remains poorly understood. Here, we determine the cryo-EM structure and activation mechanism of retron-Kva2, a type IX retron from the human pathogen Klebsiella variicola. We reveal that retron-Kva2 assembles into an asymmetric, higher-order ribonucleoprotein complex that sequesters a toxic dimeric HEPN RNase at its core. We identify a natural phage trigger as the phage T5 protein D5, which activates the retron through structural mimicry. Mirroring the retron-Kva2 winged-helix protein, the helix-turn-helix fold of D5 binds the msDNA sensor, driving conformational remodeling that unleashes HEPN-mediated tRNA cleavage and growth arrest. Because retron-Kva2 surveils a structural fold via msDNA binding, rather than a primary sequence, this recognition mechanism provides a broadly exploitable pathway for programmable activation. Harnessing this structure-based logic, we computationally designed de novo synthetic triggers that activate retron-Kva2-mediated bacterial growth arrest in vivo. Our findings reveal the architectural basis of type IX retron immunity and establish a structure-guided paradigm for repurposing bacterial defense systems into precision-honed antimicrobial therapeutics.

Advanced genome editing technologies have enabled rapid, flexible rewriting of the Escherichia coli genome, but most have not been tested in other bacterial species. Recombitrons—a genome editing tool created by pairing modified, donor-producing bacterial retrons with single-stranded binding and annealing proteins—have increased the efficiency of recombineering to install flexible, precise edits in the prokaryotic chromosome. Here, to extend their utility outside of E. coli, we surveyed the portability and versatility of retron-mediated recombineering across three different bacterial phyla (Proteobacteria, Bacillota and Actinomycetota) and a total of 15 different species. We found that retron recombineering is variable across the species tested but functional in all of them, reaching editing efficiencies >20% in six of them, >40% in three and >90% in two. Efficiencies in the remaining nine species ranged from 0.015% to 7.4%. We also tested the extension of the recombitron architecture, operon and strain modifications in a subset of hosts in which species-specific modifications were required to increase editing rates.

Diversity-generating retroelements (DGRs) are natural systems that accelerate the evolution of diverse bacterial functions through targeted hypermutation. We establish a method using DGRs coupled to recombineering (DGRec), which enables the diversification of any sequence of interest in Escherichia coli. Detailed characterization of reverse transcriptase sequence biases demonstrates how it maximizes the exploration of the sequence space while avoiding nonsense mutations. By leveraging the high error rate of the DGR reverse transcriptase at adenines, DGRec can efficiently diversify user-defined sequence windows of 50-200 bp. Mutations can be focused at specific positions, with rates reaching up to 1.38 × 10^-2 per base per generation, allowing up to 24 mutations to accumulate within a single target sequence after 48 h. We apply DGRec to phage λ host-range engineering, to the evolution of dCas9 variants and to accelerated evolution of specific nanobodies through a bacterial display setup. Lastly, we establish the feasibility of DGR-mediated mutagenesis in yeast by adapting a recombination and selection strategy previously developed for retrons.

Here, we introduce Detectrons, modular biosensors that couples programmable toehold switches with retron-mediated reverse transcription to transduce RNA inputs into unique DNA barcodes. The ability to convert dynamic RNA signals into durable DNA records within living cells unlocks powerful new modes of transcript-based sensing with applications including viral infection detection. Through the construction of a synthetic toehold-retron library and application of machine learning, we uncovered key design principles that improve signal strength and specificity. We applied Detectrons to the multiplexed live-cell detection of specific phage infections, enabling transcript-triggered barcode synthesis and quantitative host susceptibility profiling in pooled bacterial populations. Detectrons are the first RNA-to-DNA transduction system, directly linking transient RNA detection to stable, sequence-encoded DNA outputs. This platform provides a scalable and generalizable strategy for phage screening and for recording transcriptional events in complex bacterial communities.

Retrons are tripartite bacterial systems composed of non-coding RNA (ncRNA), reverse transcriptase (RT), and effector proteins with diverse enzymatic domains. Here, we characterized Retron-Eco11, a type III-A3 retron associated with a phosphoribosyltransferase-like effector, and demonstrated that it mediates anti-phage defense and can be harnessed for genome editing. All three components—ncRNA, RT, and effector—are essential for defense. Retron-Eco11 protects Escherichia coli against multiple phages and is specifically activated by structurally related, phage-encoded helicases, including UvsW and D10 from phages T4 and T5. We show that these helicases trigger effector-mediated toxicity, leading to abortive infection, which is associated with phosphoribosyl pyrophosphate depletion due to PRTase activation, thereby altering nucleotide metabolism. UvsW interacts directly with multicopy single-stranded DNA, and effector protein activation requires the catalytic activity of phage-encoded helicases. Our data indicate that activation occurs without detectable effector release or measurable changes in the composition of the Retron complex. Beyond its defensive role, Retron-Eco11 enables targeted genome editing in bacterial and eukaryotic cells, underscoring the biological relevance and biotechnological potential of type III-A3 retrons.

Retrons are bacterial immune systems that protect a bacterial population against phages by killing infected hosts. Retrons typically comprise a reverse transcriptase (RT), a template noncoding RNA that is partially reverse transcribed into RT-DNA, and a toxic effector. The reverse transcriptase (RT), noncoding RNA, and RT-DNA complex sequester the toxic effector until triggered by phage infection, at which point the toxin is released to induce cell death. Due to their ability to produce single-stranded DNA in vivo, retrons have also been engineered to produce donor templates for genome editing in both prokaryotes and eukaryotes. However, the current repertoire of experimentally characterized retrons is limited, with most retrons sourced from clinical and laboratory strains of bacteria. To better understand retron biology and natural diversity, and to expand the current toolbox of retron-based genome editors, we developed a pipeline to isolate retrons and their bacterial hosts from a variety of environmental samples. Here, we identify seven new retron systems, each isolated from a different host bacterium. We characterize DNA production by these retrons and test their ability to defend against a panel of Escherichia coli phages. We find that two of these retrons are disrupted by other elements, in one case a group II intron and in another a separate defense system, yet both retrons still produce RT-DNA. For two other retrons, we further unravel their mechanism of defense by identifying the phage genes responsible for triggering abortive infection. Finally, we engineer these retrons for genome editing in E. coli , demonstrating their potential use in a biotechnological application.