Basecamp Research unveils AI models for programmable gene insertion
The company says its EDEN platform marks a breakthrough in cell and gene therapy design, following large-scale training with NVIDIA and new investment ahead of a Series C round.
Basecamp Research has launched a new set of artificial intelligence models it says are capable of programmable gene insertion, a development the company believes could unlock a new class of cell and gene therapies.
The models are powered by EDEN, a family of evolutionary AI systems developed in collaboration with NVIDIA and trained on BaseData, Basecamp Research’s proprietary genomics dataset. The company said NVentures invested in the business ahead of a planned Series C funding round, following deep technical collaboration on the EDEN models.
Basecamp Research said laboratory results published in a peer-reviewed paper co-authored by NVIDIA, Microsoft and leading academic researchers demonstrated the capability of the EDEN models to design active insertion proteins for all tested disease-relevant target sites. According to the company, the models achieved successful designs for 100% of targets evaluated.
The company said it has demonstrated gene insertion at more than 10,000 disease-related genomic locations. This includes therapeutically relevant integration of cancer-fighting DNA into primary human T cells at novel safe-harbour sites. Basecamp Research said the resulting CAR-T cells showed more than 90% tumour-cell clearance in laboratory assays, a level of performance it described as highly encouraging for future therapeutic development.
John Finn, Chief Scientific Officer at Basecamp Research, said the results point to a step-change in what may be possible for patients. “We believe we are at the start of a major expansion of what’s possible for patients with cancer and genetic disease,” he said.
Beyond gene insertion, the same EDEN models have also been applied to antimicrobial discovery. Basecamp Research said the system designed a library of antimicrobial peptides with a 97% laboratory-confirmed activity rate. That work was carried out in collaboration with scientists at the University of Pennsylvania, led by Professor César de la Fuente.
The scale of training behind EDEN is central to the company’s claims. Basecamp Research said the models were trained on more than 10 trillion tokens of evolutionary DNA drawn from over one million newly discovered species. The data was collected over five years from more than 150 locations across 28 countries and five continents, forming what the company describes as one of the largest and most diverse biological datasets assembled for AI training.
The largest EDEN model was trained using approximately 1.95×10²⁴ floating point operations on a cluster of 1,008 NVIDIA Hopper GPUs. Training and inference were accelerated using libraries from NVIDIA BioNeMo, which is designed to support large-scale biological and chemical model development.
Basecamp Research said it plans to advance a pipeline of potentially curative therapies by continuing to expand BaseData, refine the EDEN model family and develop its aiPGI platform. The company argues that combining evolutionary-scale data with frontier AI compute could fundamentally change how gene-editing tools, cell therapies and antimicrobials are designed, shifting them from bespoke laboratory efforts to programmable, scalable systems.
The Recap
- AI models for programmable gene insertion announced by Basecamp Research.
- Trained on over 10 trillion tokens from more than one million species.
- NVentures invested ahead of Basecamp Research's forthcoming Series C round.
