News

Google's Vertex AI Can Now Run DeepMind's AlphaFold

Google's parent company, Alphabet, this week announced a new integration between its Vertex AI managed machine learning platform and AlphaFold, a protein structure prediction system developed by the company's DeepMind artificial intelligence research group.  

"We expect this capability to be a boon for data scientists and organizations of all types in the bio-pharma space, from those developing treatments for diseases to those creating new synthetic biomaterials," Shweta Maniar, director of global biopharma solutions at Google Cloud, wrote in a blog post.

Vertex AI is a managed machine learning (ML) platform designed to allow companies to accelerate the deployment and maintenance of artificial intelligence (AI) models. Introduced last spring at the Google I/O conference, platform requires nearly 80% fewer lines of code to train a model versus competitive platforms, according to Google Cloud internal research, giving data scientists and ML engineers across all levels of expertise the ability to implement Machine Learning Operations (MLOps) to build and manage ML projects throughout the development lifecycle.

AlphaFold is an AI system designed to predict three-dimensional (3D) protein structures from amino acid sequences. The DeepMind group "taught" the system to do this by showing it the sequences and structures of around 100,000 known proteins. The latest version predicts the shape of a protein at scale and in minutes, down to atomic accuracy, the company says. Google Cloud made AlphaFold available for free in July 2021.

AlphaFold is run on Vertex AI Workbench, an end-to-end, notebook-based production environment that can be preconfigured with the necessary runtime dependencies. With user-managed notebooks, devs can configure a GPU accelerator to run AlphaFold using JAX, without having to install and manage drivers or JupyterLab instances.

Google Cloud has provide a customized Docker image in its Artifact Registry with preinstalled packages for launching a notebook instance in Vertex AI Workbench, as well as prerequisites for running AlphaFold. For devs who want to further customize the docker image for the notebook instance, the company has provided the Dockerfile and a build script they can build upon. The notebook, the Dockerfile, and the build script are available in Vertex AI community content.

The company has provided a step-by-step walkthrough for launching a demonstration notebook that can predict the structure of a protein using a slightly simplified version of AlphaFold that does not use homologous protein structures or the full-sized BFD sequence database. There's also a scientific paper currently available in Nature, with lots of detail about how AlphaFold was developed and how it works.

About the Author

John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS.  He can be reached at jwaters@converge360.com.

Featured