OpenAI announced the development of a large language model named GPT-Rosalind, specifically trained on common biology workflows. The model is named after Rosalind Franklin, a prominent figure in biology, and is designed to offer a more specialized approach compared to typical science-focused models that use a generic framework.
Yunyun Wang, OpenAI’s Life Sciences Product Lead, stated the model addresses significant challenges faced by biology researchers. These challenges include managing massive datasets from decades of genome sequencing and protein biochemistry, and navigating the complexities of various specialized subfields, each with unique jargon and techniques.
For instance, a geneticist may struggle with neurobiological literature while researching brain cell-active genes. To combat these issues, OpenAI trained GPT-Rosalind on 50 of the most common biological workflows and incorporated access to major public biological databases.
The trained model can suggest likely biological pathways and prioritize potential drug targets. “We’re connecting genotype to phenotype through known pathways and regulatory mechanisms, infer likely structural or functional properties of proteins, and really leveraging this mechanistic understanding,” Wang said.
