{"id":2836,"date":"2023-01-12T15:54:27","date_gmt":"2023-01-12T15:54:27","guid":{"rendered":"https:\/\/salarydistribution.com\/machine-learning\/2023\/01\/12\/nvidia-evozyne-create-generative-ai-model-for-proteins\/"},"modified":"2023-01-12T15:54:27","modified_gmt":"2023-01-12T15:54:27","slug":"nvidia-evozyne-create-generative-ai-model-for-proteins","status":"publish","type":"post","link":"https:\/\/salarydistribution.com\/machine-learning\/2023\/01\/12\/nvidia-evozyne-create-generative-ai-model-for-proteins\/","title":{"rendered":"NVIDIA, Evozyne Create Generative AI Model for Proteins"},"content":{"rendered":"
\n

Using a pretrained AI model<\/a> from NVIDIA, startup Evozyne created two proteins with significant potential in healthcare and clean energy.<\/p>\n

A joint paper released today describes the process and the biological building blocks it produced. One aims to cure a congenital disease, another is designed to consume carbon dioxide to reduce global warming.<\/p>\n

Initial results show a new way to accelerate drug discovery and more.<\/p>\n

\u201cIt\u2019s been really encouraging that even in this first round the AI model has produced synthetic proteins as good as naturally occurring ones,\u201d said Andrew Ferguson, Evozyne\u2019s co-founder and a co-author of the paper. \u201cThat tells us it\u2019s learned nature\u2019s design rules correctly.\u201d<\/p>\n

A Transformational AI Model<\/b><\/h2>\n

Evozyne used NVIDIA\u2019s implementation of ProtT5, a transformer model<\/a> that\u2019s part of NVIDIA BioNeMo<\/a>, a software framework and service for creating AI models for healthcare.<\/p>\n

\u201cBioNeMo really gave us everything we needed to support model training and then run jobs with the model very inexpensively \u2014 we could generate millions of sequences in just a few seconds,\u201d said Ferguson, a molecular engineer working at the intersection of chemistry and machine learning.<\/p>\n

The model lies at the heart of Evovyne\u2019s process called ProT-VAE. It\u2019s a workflow that combines BioNeMo with a variational autoencoder that acts as a filter.<\/p>\n

\u201cUsing large language models combined with variational autoencoders to design proteins was not on anybody\u2019s radar just a few years ago,\u201d he said.<\/p>\n

Model Learns Nature\u2019s Ways<\/b><\/h2>\n

Like a student reading a book, NVIDIA\u2019s transformer model reads sequences of amino acids in millions of proteins. Using the same techniques neural networks employ to understand text, it learned how nature assembles these powerful building blocks of biology.<\/p>\n

The model then predicted how to assemble new proteins suited for functions Evozyne wants to address.<\/p>\n

\u201cThe technology is enabling us to do things that were pipe dreams 10 years ago,\u201d he said.<\/p>\n

A Sea of Possibilities<\/b><\/h2>\n

Machine learning helps navigate the astronomical number of possible protein sequences, then efficiently identifies the most useful ones.<\/p>\n

The traditional method of engineering proteins, called directed evolution, uses a slow, hit-or-miss approach. It typically only changes a few amino acids in sequence at a time.<\/p>\n

\"Evozyne's<\/a>
Evozyne\u2019s ProT-VAE process uses a powerful transformer model in NVIDIA BioNeMo to generate useful proteins for drug discovery and energy sustainability.<\/figcaption><\/figure>\n

By contrast, Evozyne\u2019s approach can alter half or more of the amino acids in a protein in a single round. That\u2019s the equivalent of making hundreds of mutations.<\/p>\n

\u201cWe\u2019re taking huge jumps which allows us to explore proteins never seen before that have new and useful functions,\u201d he said.<\/p>\n

Using the new process, Evozyne plans to build a range of proteins to fight diseases and climate change.<\/p>\n

Slashing Training Time, Scaling Models<\/b><\/h2>\n

\u201cNVIDIA\u2019s been an incredible partner on this work,\u201d he said.<\/p>\n

\u201cThey scaled jobs to multiple GPUs to speed up training,\u201d said Joshua Moller, a data scientist at Evozyne. \u201cWe were getting through entire datasets every minute.\u201d<\/p>\n

That reduced the time to train large AI models from months to a week. \u201cIt allowed us to train models \u2014 some with billions of trainable parameters \u2014 that just would not be possible otherwise,\u201d Ferguson said.<\/p>\n

Much More to Come<\/b><\/h2>\n

The horizon for AI-accelerated protein engineering is wide.<\/p>\n

\u201cThe field is moving incredibly quickly, and I\u2019m really excited to see what comes next,\u201d he said, noting the recent rise of diffusion models.<\/p>\n

\u201cWho knows where we will be in five years\u2019 time.\u201d<\/p>\n

Sign up for early access to the NVIDIA BioNeMo<\/a> to see how it can accelerate your applications.<\/em><\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

https:\/\/blogs.nvidia.com\/blog\/2023\/01\/12\/generative-ai-proteins-evozyne\/<\/p>\n","protected":false},"author":0,"featured_media":2837,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[3],"tags":[],"_links":{"self":[{"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/posts\/2836"}],"collection":[{"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/comments?post=2836"}],"version-history":[{"count":0,"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/posts\/2836\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/media\/2837"}],"wp:attachment":[{"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/media?parent=2836"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/categories?post=2836"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/salarydistribution.com\/machine-learning\/wp-json\/wp\/v2\/tags?post=2836"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}