OpenAI starts offering a biology-tuned LLM

On Thursday, OpenAI introduced a groundbreaking large language model designed specifically for biology, named GPT-Rosalind in honor of the renowned scientist Rosalind Franklin. This innovative model sets itself apart from other science-oriented models from major tech firms, which typically adopt a broad approach applicable across multiple disciplines. During a press briefing, Yunyun Wang, the Life Sciences Product Lead at OpenAI, explained that the model aims to address two significant challenges currently faced by biology researchers. The first challenge is the overwhelming volume of data generated from decades of genome sequencing and protein biochemistry, which can be daunting for individual researchers to navigate. The second challenge stems from the complexity of biology, which encompasses numerous specialized subfields, each with its own specific techniques and terminology. For instance, a geneticist exploring a gene relevant to brain cells may find it difficult to sift through extensive neurobiological literature. Wang noted that OpenAI has meticulously trained this LLM on 50 of the most prevalent biological workflows and equipped it with the ability to access major public databases of biological information. The result is a model capable of suggesting probable biological pathways and identifying potential drug targets. "We’re bridging genotype to phenotype through established pathways and regulatory mechanisms, and inferring likely structural or functional properties of proteins, thus leveraging this mechanistic insight," Wang added.