Basecamp Research raises $140 million to develop EDEN AI models for antibiotics and gene therapies
Basecamp Research raised $140 million from investors including S32, Nvidia, and Anthropic's Anthology Fund to develop EDEN AI models, trained on microbial DNA, for designing antibiotics and gene therapies.
London-based company Basecamp Research raised $140 million in a funding round. The round was led by investor S32, with participation from Nvidia, Anthropic's Anthology Fund, the NATO Innovation Fund, and Redalpine. The company, founded in 2020, wants to use the funds to further develop its biological AI models called EDEN and to advance its own treatment candidates into clinical development, initially focusing on gene therapies.
The EDEN models are trained on genetic material from microorganisms that the company collects around the world – from rainforests, oceans, hot springs, volcanic soil, and the deep sea near Antarctica. According to the company, its data collection network covers more than 30 countries and all seven continents; according to data from Microsoft, the project's cloud partner, 208 organizations in 31 countries participate in the collection. At the time of the interview, the dataset contained approximately 15 trillion tokens, with a token representing an individual building block of DNA – this is comparable to the scale of text datasets used to train models such as Claude or GPT. According to Microsoft, the first generation of EDEN was trained on 9.7 trillion DNA building blocks from more than one million newly sequenced species, with the computational demands of training reportedly comparable to the GPT-4 model – OpenAI has not officially confirmed this figure. According to the company, the dataset is expected to grow roughly a hundredfold over the next approximately 1.5 years and exceed one quadrillion tokens, as part of the Trillion Gene Atlas project, which also involves Anthropic, Nvidia, PacBio, and Ultima Genomics.
The company uses the EDEN models to design new antibiotics and gene therapy tools intended to insert genes at targeted locations in the body. According to the company, laboratory tests and mouse studies conducted so far show that individual drug candidates are effective against multidrug-resistant bacteria. However, clinical trials in humans still need to demonstrate safety and efficacy. CTO Philip Lorenz told The Decoder that biology is a much more complex problem for AI than language, and that good results on paper do not guarantee functional molecules; he also noted that public genomic databases present an incomplete picture of nature, since approximately 68 percent of the sequence volume in the Sequence Read Archive database comes from just five species, with humans alone accounting for about 54 percent.
The source article was not available in full; further details can be found in the source article.
Why it matters
For the drug development field, this is an example of a specialized biological AI model trained on a proprietary, large-scale genomic dataset instead of limited public databases, which the company says enables broader coverage of organisms than existing sources. Investment by major technology players (Nvidia, Anthropic) in this approach shows interest in the convergence of infrastructure for training both language and biological models. Results so far are limited to laboratory and mouse models, so clinical trials in humans still need to demonstrate the real benefit for patients.
Relevant practical impact
What this means
For a business
Investment by major technology players (Nvidia, Anthropic's Anthology Fund) in specialized biological AI models signals growing interest in domain-specific models outside the language domain, which is relevant context for companies considering partnerships or investments at the intersection of AI and biotechnology; however, the commercial and clinical benefit is not yet proven.
DevelopmentCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.