Huggins Face released Carbon-X, a 14B-parameter open-source model combining genomic sequences and natural language to query DNA changes.
Running locally without sending data externally, it excludes virus sequences and runs on user hardware, though its predictions await wet lab validation.
The new model, called Carbon-X, was created in collaboration with researchers from Zhongguancun Academy in China and the University of Naples Federico II in Italy. The tool combines genetic sequences and natural language, reading DNA in groups of six letters.
According to senior researcher Lewis Tunstall, the aim of the tool is to help scientists design drugs and genetic therapies, and it may also help patients.
How the system was trained
Carbon-X was trained on 300 billion data chunks, combining genetic sequences from a variety of organisms—including mammals, vertebrates, invertebrates, plants, fungi, and protozoa—along with English text. The creators intentionally excluded viruses from the training data to prevent potentially dangerous applications.
READ ALSO
Anthropic introduces Claude Duet, a new digital assistant for professionals
Users cannot upload the entire human genome, as the model processes sections of genetic material of about 50.000 base pairs at a time.
Local operation and warnings
The model has 14 billion parameters and is designed to run locally on the user's computer, without sending data to Hugging Face or other companies.
However, the creators warn that Carbon-X has not yet been verified with real laboratory experiments, so the accuracy of its predictions remains unverified.
G
Hacks.gr on Google
Set it as a preferred source for cybersecurity updates.
Preferred source
Top comments (0)