13 January 2026

 

Basecamp Research, a frontier AI lab operating between London and Cambridge, Massachusetts, says it has reached a long-sought milestone in genetic medicine: using artificial intelligence to design enzymes capable of inserting large DNA sequences at precise locations in the human genome. If the approach proves robust beyond the laboratory, it could significantly expand the scope of cell and gene therapies for cancer and inherited diseases.

The company announced this week that it has developed what it describes as the first AI models able to perform programmable gene insertion. Built in close technical collaboration with NVIDIA, the models underpin a new platform called AI-Programmable Gene Insertion (aiPGI™). Alongside the scientific announcement, Basecamp Research disclosed an investment from NVentures, NVIDIA’s venture capital arm, as part of its pre-Series C funding round.

According to Basecamp Research, the work addresses a core limitation in current gene-editing technologies and opens a pathway toward more predictable and precisely engineered therapies.

Moving beyond the limits of CRISPR

Programmable gene insertion—the targeted placement of large therapeutic DNA sequences into the genome—has been a central ambition in genetic medicine for decades. While CRISPR-Cas systems have transformed genome editing, they remain constrained. Most CRISPR-based approaches rely on inducing breaks in DNA and typically enable only small edits, limiting both where changes can be made and how reliably they can be controlled.

These constraints matter clinically. DNA damage can introduce variability, off-target effects, and safety concerns, particularly when editing cells intended for therapeutic use. As a result, many promising gene-therapy concepts have remained difficult to translate into predictable, scalable treatments.

Basecamp Research says its aiPGI™ platform offers an alternative. Rather than modifying existing enzymes, the company uses AI to design entirely new insertion proteins capable of placing large DNA payloads at defined genomic locations—without relying on double-strand DNA breaks.

“We believe we are at the start of a major expansion of what’s possible for patients with cancer and genetic disease,” said John Finn, chief scientific officer at Basecamp Research. “By using AI to design the therapeutic enzyme, we aim to accelerate the development of treatments for diseases that have remained out of reach.”

EDEN: evolutionary AI trained on global genomic diversity

The aiPGI™ platform is powered by EDEN, a family of evolutionary AI models developed with NVIDIA. EDEN was trained on BaseData™, Basecamp Research’s proprietary genomics dataset, which the company describes as the largest of its kind.

BaseData™ contains more than 10 trillion tokens of evolutionary DNA derived from over one million newly identified species. The data was collected over five years from more than 150 sites across 28 countries and five continents, using a data-collection strategy designed to capture genetic diversity that is poorly represented in public databases. Basecamp published details of this approach in 2025, positioning ethical sourcing and global representation as foundational to its model development.

The scale of computation involved reflects the growing convergence between frontier AI and life sciences. The largest EDEN model was trained using a cluster of 1,008 NVIDIA Hopper GPUs and nearly 2 × 10²⁴ floating-point operations, placing it in the same computational class as leading general-purpose AI systems. Training was accelerated using NVIDIA’s BioNeMo libraries, which are increasingly used for large-scale biological modeling.

In a peer-reviewed paper published this week and co-authored with researchers from NVIDIA, Microsoft and academic institutions, Basecamp reports that EDEN designed active insertion proteins for all tested disease-relevant genomic target sites. The models required only the DNA target site as input, a capability the authors describe as a meaningful step forward in AI-driven biological design.

Early laboratory results in cancer cell therapy

Beyond model performance, Basecamp Research has also reported early experimental results. The company says it has demonstrated successful insertion at more than 10,000 disease-related locations in the human genome.

One highlighted application involves engineering primary human T cells. Using aiPGI™, Basecamp integrated cancer-fighting DNA into novel “safe-harbour” sites—genomic regions considered suitable for therapeutic insertion—rather than commonly used locations with known trade-offs. The resulting CAR-T cells showed more than 90% tumour-cell clearance in laboratory assays.

While these findings remain preclinical, they suggest a potential route to more predictable and customisable cell therapies. CAR-T treatments are already used clinically for certain cancers, but current manufacturing approaches remain complex, costly and variable.

Applying the same models to antimicrobial resistance

Basecamp Research also used EDEN to tackle a different global health challenge: antimicrobial resistance. The World Health Organization has repeatedly warned that drug-resistant infections threaten to undermine decades of medical progress, with limited new antibiotics entering the market.

In collaboration with researchers at the University of Pennsylvania led by Professor César de la Fuente, Basecamp applied its AI models to design antimicrobial peptides (AMPs)—small proteins that can kill bacteria and are seen as promising alternatives to conventional antibiotics.

The company reports that 97% of the AI-designed AMP candidates demonstrated confirmed activity in laboratory testing. The strongest performers showed high potency against critical-priority, multidrug-resistant pathogens. While still early-stage, the results highlight how generative biological models can be applied across multiple therapeutic domains.

From proprietary data to therapeutic pipelines

Founded in the UK, Basecamp Research positions itself around the idea of exploring “Beyond Known Biology™”—using AI to access regions of biological sequence space that are absent from public reference datasets. To do so, the company collects and curates its own biological data through partnerships with more than 150 organisations worldwide, including research institutions and conservation groups.

This strategy has helped Basecamp stand out in a crowded AI-biotech landscape. The company has been recognised by Fast Company as one of the most innovative biotech firms and named to the FT-backed Sifted AI100 list of Europe’s leading AI startups.

Its business model combines internal therapeutic development with partnerships across biopharma and academia. The capabilities demonstrated with aiPGI™ now underpin an emerging pipeline of cell and gene therapies, with the company aiming to develop more precise and personalised treatments across cancer and genetic disease indications.

The new investment from NVentures formalises a collaboration that has already spanned several years. For NVIDIA, which has increasingly positioned itself as a platform provider for scientific and biomedical AI, the partnership underscores the strategic importance of life sciences as a frontier market for large-scale computation.

AI’s expanding role in the biosciences

Basecamp Research’s announcement reflects a broader shift underway in the biosciences. Over the past decade, AI has moved from a supporting role into core scientific infrastructure, transforming protein structure prediction, drug discovery and systems biology.

High-profile breakthroughs such as AlphaFold have demonstrated how machine learning can solve problems that resisted experimental approaches for decades. Today, most large pharmaceutical companies use AI for target identification, molecular design and clinical development, while regulators such as the US Food and Drug Administration have begun publishing guidance on the use of machine learning in medical products.

Attention is now increasingly focused on generative models that can design biological components rather than merely analyse them. At the same time, gene editing remains among the most tightly scrutinised applications, with ongoing debates around safety, delivery and long-term effects.

Whether AI-designed programmable gene insertion can move from laboratory proof-of-concept to clinical reality remains an open question. What is clear, however, is that biology is becoming one of the most computationally intensive and strategically important domains in AI.

For Basecamp Research, the latest results mark a significant technical milestone. For the wider field, they offer another signal that the boundary between artificial intelligence and biological engineering is rapidly dissolving—reshaping how new medicines may be discovered and built.

Find out more: How AI models “learn” biology without understanding it

 

If you value independent journalism on AI and deep tech, you can support our work via our page on Buy Me A Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.