Google DeepMind Announcing the launch of AlphaProteo, an AI system that helps biology and health researchers design novel high-potency proteins that bind precisely and strongly to target molecules.
AlphaProteo is trained on the Protein Data Bank (PDB), which helps advance breakthroughs in science and education by providing access and tools to explore, visualize, and analyze experimentally determined 3D structures from the PDB archive.
Based on the structure of the target molecule and its set of preferred binding sites, AlphaProteo generates candidate proteins that bind to the target.
The tech giant said the binder has the potential to open up new areas of research in drug development and diagnostic biosensors.
“AlphaProteo can generate novel protein binders for a variety of target proteins, including VEGF-A, which has been linked to cancer and diabetes complications. This is the first time that an AI tool has been able to successfully design a protein binder for VEGF-A,” Google DeepMind’s Protein Design and Wet Lab teams said in a blog post.
“AlphaProteo also achieved higher experimental success rates and 3- to 300-fold better binding affinities than the best existing methods for seven target proteins we tested.”
To test AlphaProteo, the AI developers designed binders for a variety of target proteins, including “two viral proteins involved in infection, BHRF1 and the SARS-CoV-2 spike protein receptor binding domain, SC2RBD; and five proteins involved in cancer, inflammation, and autoimmune diseases: IL-7Rɑ, PD-L1, TrkA, IL-17A, and VEGF-A.”
The binding success rate for one viral target, BHRF1, was 88% on average, which was 10 times higher than that of conventional methods.
The Google DeepMind Web Labs team, working with external research groups including researchers at the Francis Crick Institute, confirmed data showing that the AlphaProteo binder blocks SARS-CoV-2 from infecting human cells.
AlphaProteo has demonstrated that it can reduce the time required for initial experiments involving protein binders for a variety of applications.
But despite these breakthroughs, researchers point out that AI systems have limitations.
For example, AlphaProteo failed to produce a successful binder for TNFa, a protein associated with autoimmune diseases such as rheumatoid arthritis.
“We chose TNFɑ to strongly challenge AlphaProteo because computational analysis showed that designing a binder for it would be extremely difficult. We will continue to improve and expand AlphaProteo’s capabilities to ultimately tackle such a challenging target,” the authors write.
The AlphaProteo research team plans to collaborate with the scientific community to further understand its limitations by observing how AlphaProteo impacts other biological problems.
Additionally, the research team has been studying the use of drug design at Isomorphic Labs.
The bigger trend
In June, Google Research and Google DeepMind Publish a paper We announced the creation of a new LLM for drug discovery and therapeutic development, called Tx-LLM, which is detailed in Med-PaLM 2.
of tech giants Med-PaLM 2 is a generative AI technology that leverages Google’s LLM to answer medical questions.
In May According to a study conducted by Google Research in collaboration with Google DeepMind, the tech giant has scaled the performance of its AI models for Med-Gemini-2D, Med-Gemini-3D, and Med-Gemini Polygenic.
Google says it fine-tuned Med-Gemini’s capabilities using histopathology, dermatology, 2D and 3D radiology, genomics, and ophthalmology data.
In 2023, Google MedLM is built on top of Med-PaLM 2 with two base models designed to answer medical questions, generate insights from unstructured data, and summarize medical information.
The company says that when it piloted LLM in healthcare organizations, it found that the most effective AI models were those designed to solve specific use cases.
As a result, MedLM’s large-scale model is built to handle complex tasks, while the other is a medium-sized model that can be fine-tuned and scaled across a variety of tasks.
The HIMSS Healthcare Cybersecurity Forum will be held in Washington, DC from October 31 to November 1. Learn more and register.
