How LLMs and Generative AI are Revolutionizing AI for Science with Anima Anandkumar - 614

Protein Folding

The advancements in protein folding using generative AI models are transforming biological research. Anima Anandkumar highlights how these models not only predict the static 3D structure of proteins but also their dynamic changes when binding to small molecules 1. This capability enhances our understanding of protein functions, potentially leading to the creation of better drugs and therapeutic targets. She explains, "We are solving all these structure problems. We can speed up our prediction of structure, we can greatly increase the accuracy of our structure prediction. But what about function?" 1. The integration of AI in drug discovery is also emphasized, where generative models can extrapolate beyond existing data to design novel molecules 2.

   

Genome Modeling

Language models are being applied to genome sequences, offering new insights into biological processes. Anima discusses the development of a 25 billion parameter biological language model that learns from genome sequences to predict new gene variants 3. This approach allows for the generation of new gene sequences and proteins, advancing our understanding of life's code. She notes, "We need to understand the language of the genome and not only learn those relationships between genes, because they map to proteins, but also potentially generate new gene sequences and through that, new proteins and new targets" 4. This work is paving the way for breakthroughs in understanding the functional implications of genetic material.