In an extraordinary leap forward for the field of genetic engineering, researchers at Integra Therapeutics, in collaboration with the Pompeu Fabra University (UPF) Department of Medicine and Life Sciences and the Center for Genomic Regulation (CRG), have showcased a groundbreaking approach to enhance the capabilities of genome editing tools. Their landmark work, recently published in the prestigious journal Nature Biotechnology, unveils the discovery and design of synthetic proteins that surpass natural counterparts in efficiency, fundamentally reshaping how therapeutic genes can be introduced into human cells with greater precision and effectiveness.
At the core of this breakthrough lies the sophisticated use of transposases—enzymes capable of cutting and pasting DNA sequences within genomes. Among these enzymes, PiggyBac transposases have gained prominence for their potential in gene therapy applications, notably in inserting therapeutic DNA segments safely into patient cells. Despite their promise, natural and previously engineered PiggyBac transposases have confronted inherent limitations, including a lack of diversity and reduced precision. Addressing these constraints, the scientific team leveraged a two-pronged strategy: extensive exploitation of biological diversity and the cutting-edge application of generative artificial intelligence (AI) to not only explore but surpass nature’s design.
The journey began with an immense computational bioprospecting effort, systematically screening over 31,000 eukaryotic genomes to unearth novel PiggyBac sequences. This ambitious endeavor revealed more than 13,000 previously unknown transposase variants harboring immense untapped potential. Subsequent experimental assays in cultured human cells confirmed the activity of 10 of these transposases, highlighting significant functional diversity that had remained hidden in natural biological reservoirs. Strikingly, among these newly discovered enzymes, two displayed performance metrics on par with the best laboratory-optimized counterparts, with one demonstrating robust activity in primary human T cells, a cell type pivotal for cutting-edge cancer immunotherapies.
While mining biodiversity expanded the known repertoire of these enzymes, the true horizon-pushing advance came next: the use of protein large language models (pLLMs), a class of generative AI. By training these sophisticated models on the vast dataset of newly discovered PiggyBac sequences, researchers culminated in the design of entirely novel synthetic transposases with enhanced activity profiles. This represents a paradigmatic shift—where AI-driven protein design transcends the evolutionary constraints of natural sequences to forge molecules tailored for superior therapeutic utility. The study confirmed that the AI-engineered transposases are not only hyperactive but seamlessly compatible with existing advanced gene editing platforms, such as Integra Therapeutics’ proprietary FiCAT system, which integrates precise DNA insertion capabilities.
Integra Therapeutics’ CEO, Dr. Avencia Sánchez-Mejías, underscores the transformative potential of this innovation, emphasizing that their findings open new avenues for revolutionizing gene editing applications and solidify the company’s leadership role in the burgeoning gene therapy landscape. This fusion of synthetic biology with AI-driven design exemplifies an era where computational tools and wet-lab experiments synergize to create next-generation therapeutic modalities with unprecedented precision and efficiency.
The underlying technology capitalizes on the remarkable ability of protein language models to decode and generate protein sequences analogous to natural languages. Dr. Marc Güell, scientific director at Integra Therapeutics and ICREA researcher at UPF, draws an analogy to ChatGPT’s text generation prowess, illustrating how these AI systems internalize the “grammar” of protein sequences—learning patterns that govern structural stability and functional properties. By speaking this language fluently, the pLLMs can conceive novel proteins maintaining both biological validity and enhanced functional characteristics, representing a genuine leap toward de novo protein engineering.
Further emphasizing the interdisciplinary nature of this research, Dr. Noelia Ferruz from the CRG’s Artificial Intelligence for Protein Design Group highlights that training these models on the entirety of known protein sequences unlocks insights inaccessible to conventional methodologies. The AI’s capacity to infer and generate sequences aligned with physical and chemical constraints enables the design of enzymes that nature has simply not discovered or optimized, significantly expanding the toolbox available for precision medicine.
This milestone achievement not only showcases the power of combining biodiversity exploration with generative AI but also addresses critical bottlenecks in the development of cellular therapies. For instance, the newly designed hyperactive PiggyBac transposases can facilitate the efficient engineering of CAR-T cells, a frontline treatment modality for certain cancers, thereby improving the transduction efficiency and potentially reducing manufacturing costs and timelines.
Moreover, this approach holds transformative implications for treating rare genetic diseases, where precise and safe insertion of large therapeutic genes remains a formidable challenge. The enhanced activity and expanded target range of synthetic transposases hold promise for creating more effective gene therapies, increasing the likelihood of successful clinical outcomes and expanding patient access to life-altering treatments.
The application of AI in this context symbolizes a broader trend in biotechnology where machine learning algorithms complement biological insights to accelerate innovation cycles. By pushing beyond natural evolutionary constraints, researchers can rationally design molecular tools tuned for specific clinical and research needs, heralding a new chapter in synthetic biology and personalized medicine.
As Integra Therapeutics continues to refine and expand the FiCAT platform augmented by these AI-designed transposases, strategic collaborations with leading academic institutions like UPF and CRG remain integral. Such partnerships accelerate the translation of fundamental research into therapeutic realities, bridging computational biology, molecular engineering, and clinical applications.
In sum, this pioneering study offers a vivid glimpse into the future of gene editing—a future where the frontiers of natural biodiversity meld with AI-generated innovation to craft superior genomic scissors. These advances promise not just incremental but exponential improvements in the development of gene and cell therapies, potentially revolutionizing treatments for cancer, rare diseases, and beyond.
Subject of Research: Not applicable
Article Title: Discovery and protein language model-guided design of hyperactive transposases
News Publication Date: 2-Oct-2025
Web References: https://integra-tx.com/, https://www.upf.edu/web/biomed/inici, https://www.crg.eu/, https://synbio.upf.edu/, https://www.crg.eu/en/programmes-groups/ferruz-lab
References: DOI 10.1038/s41587-025-02816-4
Image Credits: Integra Therapeutics
Keywords: Synthetic biology, Artificial intelligence, Personalized medicine
Tags: breakthroughs in synthetic biologycollaboration in biotechnology researchcomputational bioprospecting in geneticsenhancing genome editing toolsgenerative artificial intelligence in protein designgenome editing advancementsinnovative approaches in genetic modificationovercoming limitations of natural enzymesPiggyBac transposases applicationsprecision in therapeutic gene introductionsynthetic proteins for gene therapytransposases in genetic engineering
