Product reviews, deals and the latest tech news

With BioNeMo, Nvidia advances generative AI for biology

Nvidia announced the debut of its BioNeMo Large Language Model (LLM) service in September 2022 with the goal of assisting academics in developing novel AI models in the biological sciences.

Nvidia and biotech firm Evozyne have announced today that they have utilised BioNeMo to create a new generative AI model that may have far-reaching effects in the areas of human health and climate change. Two novel proteins, both of which were designed using the generative AI paradigm, are being described here. Potential applications for these proteins include sequestering carbon dioxide and treating congenital disorders, respectively.

“Proteins are the building elements of life,” Kimberly Powell, Nvidia’s vice president of healthcare, said during a press conference. Discovering novel proteins that can aid in medicine design, carbon capture, and eco-friendly textile production is a primary goal of protein engineering.

A revolution in healthcare is underway, and generative AI is at the forefront.

ChatGPG’s popularity over the last several months has made generative AI a hot subject of conversation, but its potential impact on the life sciences is really revolutionary.

This week, life science company Absci revealed it has developed a generative AI model utilising E. coli bacteria to develop novel antibody medicines. To aid with these endeavours, IBM has created a free library called the Generative Toolkit for Scientific Discovery (GT4SD).

Overall, in 2023, researchers will likely focus on using generative AI to speed up drug development in an effort to shorten the time it takes to develop new therapies.

Implementing Generative AI with Nvidia BioNeMo

For instance, Nvidia has developed a model called the BioNeMo LLM specifically to facilitate drug development.

Generative AI is an approach to AI that allows for the creation of novel content by using a pre-trained model of a huge body of language to generate ideas.

According to Powell, there are infinitely more possible proteins than there are atoms in the cosmos. As a result, Evozyne and Nvidia utilised Nvidia BioNeMo to rapidly develop a predictive model for protein discovery called Prot-VAE.

She said that Evozyne has utilised BioNeMo to tailor Prot-VAE to a particular class of proteins. Once the new model was optimised, it was utilised to create a new library of proteins that either preserve or improve a function already present in a predefined family of proteins.