For example, computational approaches led by the Materials Project and other groups have contributed to the development of 28,000 new materials to date.
But this is an expensive and time-consuming process, and researchers can find it difficult to develop radically different structures because they usually work with existing materials.
However, recently, Google unit DeepMind Technologies' AI tool called Graph Networks for Materials Exploration (GNoME) helped create 2.2 million crystals, of which 380,000 are stable materials.
This is significant progress as it could help researchers develop greener technologies such as more efficient batteries for electric cars, solar photovoltaics, superconductors and more efficient computing.
GNoME is a deep-learning tool that predicts the stability of new materials, increasing the speed and efficiency of discovery and allowing researchers to create materials faster and at scale. Google says its new discovery is “equivalent to nearly 800 years of knowledge.”
GNoME's predictions are accessible to scientists around the world. A team of researchers from Lawrence Berkeley National Laboratory, in collaboration with DeepMind, has published a paper revealing how AI predictions can be used for autonomous materials synthesis. The laboratory uses machine learning and robotic arms to create new materials.
How does this all scale up?
Other researchers have independently created 736 new materials from GNoME in their labs, according to Google. The AI tool has identified 52,000 new layered compounds, similar to graphene, that can be used instead of silicon to make superconductors. About 1,000 such materials had previously been identified.
GNoME also found 528 potential lithium-ion conductors, 25 times more than a previous attempt, that could be used to improve the performance of rechargeable batteries.
Google describes GNoME as a graph neural network model (GNN), where the input data takes the form of a graph. The AI model was originally trained using data on crystal structures and their stability, which was openly available through the Materials Project.
Google continuously assessed the performance of its model using computational techniques known as Density Functional Theory (DFT) – used in physics, chemistry and materials science to understand the structures of atoms – to assess the stability of crystals. It also used a training process called 'active learning' that improved GNoME's performance.
GNoME would generate predictions for the structures of new, stable crystals, which were then tested using DFT. The resulting training data was then fed back to the training model.
“Our research increased the discovery rate of material stability prediction from approximately 50% to 80% based on MatBench Discovery, an external benchmark established by previous state-of-the-art models,” Google said in a blog post.
“We also managed to scale the efficiency of our model by improving the discovery rate from less than 10% to more than 80%. Such efficiencies could have a significant impact on the amount of computing power required per discovery.”
That said, the AI models have limitations. For example, Berkeley's A-Lab failed to create 17 of the 58 materials it targeted. Some materials required heating to higher temperatures or better grinding, which are standard steps in labs outside the current scope of AI.
Experts also point out that, similar to other AI systems, the models don't explain how they reached their decision, which won't help if other researchers want to understand the process.
Can AI solve the mystery of life?
Only in July 2022, DeepMind Technologies released 3D predicted structures of more than 200 million proteins found in plants, bacteria, animals and humans. The structures, generated by DeepMind's AI system, AlphaFold, will help researchers advance their understanding of how proteins fold.
This discovery was critical because proteins are the building blocks of human life, along with nucleic acids (DNA and RNA), lipids and glycans, and folding allows a protein to adopt a functional shape or conformation.
If researchers can better predict how proteins fold, they will be able to better understand how cells function and how misfolded proteins can cause diseases, known as the protein folding problem.
AlphaFold uses an AI system to predict the 3D structure of a protein based on its amino acid chain. Announced in 2020, DeepMind released and open sourced AlphaFold2, and in 2021 the multi-terabyte AlphaFold Protein Structure Database (AlphaFold DB, which the company likens to a 'Google search' for protein structures).
A year later, DeepMind, together with EMBL's European Bioinformatics Institute (EMBL-EBI), published predicted structures for more than 200 million proteins, covering almost every cataloged protein in science. “One day it could even help unravel the mysteries of how life itself works,” according to a Google blog.
The importance of achieving atomic accuracy
According to Google, 1.4 million users in at least 190 countries have accessed the AlphaFold database to date. Scientists around the world have used AlphaFold's predictions to advance research, from accelerating new malaria vaccines and advancing cancer drug discovery to developing plastic-eating enzymes to tackle pollution.
The University of Portsmouth's Center for Enzyme Innovation is using the AI system to help develop faster enzymes to recycle single-use plastics, while the University of California has used the predictions to understand the biology of the Covid-19 (SARS-CoV ) easier to understand. -2) virus.
AlphaFold's latest model can generate predictions for almost all molecules in the Protein Data Bank, “often achieving atomic accuracy.”
Levinthal's paradox, named after molecular biologist Cyrus Levinthal, notes that even if proteins could fold in seconds or even milliseconds, it would take longer than the age of the known universe (about 13.8 billion years) to complete all possible configurations. of a typical protein using brute force.
AI systems such as AlphaFold and RoseTTAFold are leveraging advances in the application of AI to dramatically improve the way drugs are discovered and developed. AlphaFold can do in seconds “what used to take many months or years,” said Eric Topol, founder and director of the Scripps Research Translational Institute.