Project EDITOME (Engineering from Data-driven Insights for Targeted Optimization of Microbiome Editing) aims to pioneer a transformative approach to genetic engineering by editing microbial genes directly within their complex, native environments. While traditional microbiome interventions rely on broad, blunt instruments like antibiotics or probiotics, EDITOME operates precisely at the genetic level, unlocking a level of precision and control that hasn’t been possible before. By utilizing proprietary Large Language Models (LLMs), the project aims to design DNA elements that can quickly and predictably change their behavior, offering unprecedented control over their growth and functionality, including in species that have been difficult or impossible to engineer. This technology is designed to address a central bottleneck in microbiome engineering: delivering tools accurately to specific target microbes within dense ecosystems and ensuring those tools function reliably. Ultimately, EDITOME could help move microbiome research from mere correlation to true causation, creating scalable, programmable interventions across human health, agriculture, and industrial sustainability.

Traditional methods of modifying microbial communities remain limited by a basic challenge: microbes often behave differently in isolation than they do within their native communities. Many methods require isolating individual strains, culturing them in a laboratory setting, applying genetic edits, and reintroducing them into their original ecosystems—a sequence that is slow, error-prone, and frequently fails because the engineered microbes may not survive or behave as expected once returned to their native environments. At the same time, many therapeutic and agricultural interventions, including antibiotics and probiotics, affect microbial communities broadly rather than making precise, targeted changes. The fundamental bottleneck holding back synthetic biology today is the inability to deliver gene-editing tools accurately to the right microbes within dense, complex living systems and getting them to execute reliably.

EDITOME introduces an AI-powered "Design-Build-Test-Learn" cycle in which computational models design genetic tools, lab experiments test them, and the results are fed back to improve the next round of designs:

• Generative LLM for Plasmids: Instead of manual trial-and-error, the team is building an LLM with the ultimate goal of generating complete, custom plasmid editing systems tailored to any target organism. As a foundational step, the model currently focuses on designing DNA sequences for compatible replicons—the critical genetic elements that allow a plasmid to persist and function inside a specific bacterium.


• ET-Seq, DART, and Bridge Recombinase Technology: Utilizing foundational IGI technologies—Environmental Transformation Sequencing (ET-Seq), which identifies microbes that can be genetically manipulated within complex communities, and DNA-editing All-in-one RNA-guided CRISPR-Cas Transposase (DART), and more recently Bridge Recombinase, which enables targeted DNA insertion. Together, these tools make it possible to edit and validate genetic changes directly within complex microbial communities, without isolating individual species.


• Iterative Learning: Experimental results from the lab are fed back into the AI model, creating a self-improving system that increases the success rate of editing bacteria that have previously been difficult to engineer.

EDITOME employs a high-throughput “Design-Build-Test-Learn” cycle that integrates generative AI with advanced CRISPR-based community editing. The approach moves beyond traditional trial-and-error by predicting the genetic elements needed to make editing tools work inside specific microbes.

The foundation of the project is a Generative Large Language Model (LLM) trained on massive genomic and plasmid databases (such as PLSDB and IMG/PR).

The Focus: While the overarching vision is end-to-end plasmid generation, the model's immediate focus is designing compatible replicons—the DNA elements, including origins of replication and Rep proteins, that allow a DNA plasmid to exist and function inside a target bacterium.

Precision Engineering: Instead of using generic tools, the AI designs custom DNA sequences optimized for the specific cellular machinery of non-model microbes that are difficult to engineer with existing methods.

In the near term, the project is focusing on industrial microbial systems, representing the lowest technical risk. This tier aims to optimize native microbial strains already utilized in manufacturing processes, such as fermentation, enzyme production, and bio-based chemicals. By surgically improving the yield, stability, and efficiency of these industrial strains without requiring costly, full-scale strain redevelopment, EDITOME could provide a highly feasible path to value within 12 to 18 months using controlled in vitro and ex vivo systems.

Moving into the medium term, the platform could expand into agriculture and climate sustainability to address broader ecological challenges. This tier leverages in situ editing to modify soil and plant-associated microbiomes, directly improving crop yields, nutrient utilization, and environmental resilience. The same approach could also be applied to livestock microbiomes, with the goal of reducing methane emissions, and to native microbial communities involved in biofuel production. Validating these applications requires an estimated 18 to 24 months and incorporates animal, field, or soil studies supported by external Contract Research Organizations (CROs).

Human health and localized therapeutics represent one of EDITOME’s highest-impact application areas. Health-focused research is already underway in laboratory models, where EDITOME can be used to test the causal role of microbial genes that were previously only correlated with human health and disease, including in areas such as IBD and asthma. By editing these genes directly and observing the resulting effects, the platform can help move microbiome research from association toward mechanism, while also supporting early demonstration of candidate treatments in controlled models. Full clinical translation remains a longer-term goal, with initial directions likely to focus on more accessible sites, such as the skin and oral cavity, where delivery is more straightforward, followed by extensive validation in living systems over longer timelines.