The latest version of the AI 'AlphaFold' that accurately predicts the three-dimensional structure of proteins has been released, making it possible to predict molecules with higher accuracy than before
In 2018,
A glimpse of the next generation of AlphaFold - Google DeepMind
https://deepmind.google/discover/blog/a-glimpse-of-the-next-generation-of-alphafold/
DeepMind's latest AlphaFold model is more useful for drug discovery | TechCrunch
Google DeepMind boss hits back at Meta AI chief over 'fearmongering' claim
https://www.cnbc.com/2023/10/31/google-deepmind-boss-hits-back-at-meta-ai-chief-over-fearmongering-claim.html
Google Brain founder says big tech is lying about AI extinction danger
https://www.afr.com/technology/google-brain-founder-says-big-tech-is-lying-about-ai-human-extinction-danger-20231027-p5efnz
Proteins, which are involved in all biological processes such as muscle contraction, blood transport, light sensing, and food energy conversion, are three-dimensional polymeric compounds made up of as many as 20 types of L-amino acids linked together in chains. To date, humans have discovered more than 200 million types of proteins, but only a small number of proteins have had their three-dimensional structures identified.
However, AlphaFold, developed by Google DeepMind, uses AI to predict the structure of proteins whose three-dimensional structure was previously unknown, and in 2020, AlphaFold predicts the structure of proteins using conventional cryo-electron microscopy and other experimental methods. It has been reported that the analysis of the three-dimensional structure of
AlphaFold is expected to have the potential to solve the so-called ' protein folding problem ,' which has been a major biological problem for many years and makes it difficult to predict three-dimensional structures from amino acid sequences. .
How is the AI 'AlphaFold' that predicts the three-dimensional structure of proteins changing the world of biology? -GIGAZINE
So far, Google DeepMind has released AlphaFold2 , an improved version of the previous AlphaFold, in 2020, and open sourced AlphaFold in 2021.
The latest version of AlphaFold, released on October 31, 2023, is now able to predict the three-dimensional structures of almost all molecules covered in the ' Protein Structure Data Bank ,' the world's largest open-access database of biomolecules. It has been reported.
Excited to share major progress on next generation of AlphaFold! Together with @IsomorphicLabs , we've significantly advanced accuracy & expanded coverage beyond proteins to other key biomolecular classes. Continuing to accelerate science at digital speed https://t.co/zmR06BlrQx
— Demis Hassabis (@demishassabis) October 31, 2023
In addition, according to Google DeepMind, the new AlphaFold also uses substances called 'ligands' that form complexes with biomolecules to serve biological purposes, as well as molecular structures associated with chemical changes that occur after proteins are made. It can be predicted accurately.
Until now, pharmaceutical researchers have used computer simulations called ``docking methods'' to study how proteins and ligands interact. However, the structure of the protein is important in the docking method, and there are limits to pharmaceutical technology using the docking method. With the new AlphaFold, AI automatically predicts protein structures, etc., so it is possible to predict proteins whose three-dimensional structures have not been revealed so far, and at the same time predict how proteins and nucleic acids interact with other molecules. It is possible to perform detailed simulations of what will happen.
The demo video of AlphaFold released by Google DeepMind is below.
Series of predicted structures compared to ground truth (white) from our latest AlphaFold model. - YouTube
Google DeepMind says, ``The new AlphaFold significantly outperforms the previous AlphaFold on various protein structure prediction problems related to drug discovery, such as antibody binding. 'This has the potential to significantly advance our scientific understanding of the molecules that make up the human body, demonstrating the utility of AI.'
On the other hand, Google DeepMind has reported (PDF file) that AlphaFold is inferior to conventional methods in predicting the structure of RNA molecules in the body, and has revealed that it is conducting further research to find a solution. I am. Additionally, Isomorphic Labs, an affiliate of Google DeepMind, has already implemented the new AlphaFold into therapeutic drug design and has reported that AlphaFold is useful for analyzing various molecular structures important for disease treatment. I am.
Demis Hassabis, CEO of Google DeepMind, predicts that an artificial intelligence called ``general purpose artificial intelligence'' that can outperform humans in various fields will eventually be born, and he predicts that AI will run out of control and will lead to crimes. We are concerned about the dangers of artificial intelligence being used, and are calling for the urgent introduction of regulations on artificial intelligence.
On the other hand, Andrew Ng, co-founder of Google's artificial intelligence development team Google Brain, said, ``The idea that artificial intelligence will wipe out humans was due to the expectation that regulations would be introduced to avoid competition in the AI market.'' This is a ``bad idea'' by major AI companies.'' ``This idea has the potential to kill new innovations in AI.''
Related Posts:
