Highlights:
- The latest AlphaFold version from Google DeepMind boasts improved predictive abilities for determining the shapes of various biological molecules.
- The enhanced version of AlphaFold is well-suited for tasks like predicting the structure of substances known as ligands, among others.
A new iteration of AlphaFold, an AI model developed by Google DeepMind to aid scientists in studying biological molecules like proteins, was recently made public.
According to the Google LLC division, the most recent version of the model offers more precise information about the structure of proteins than the one before. Moreover, AlphaFold can now research a wider variety of biological molecules. DeepMind thinks that the improved capabilities of the model could contribute to the advancement of research in fields like drug discovery.
The building blocks of life, proteins, are intricate molecules that can fold and twist into various shapes. A protein’s configuration directly impacts the structures it behaves in. For this reason, scientists have made it a top priority to study those structures.
Sometimes, it can take years to determine a protein’s shape manually. Scientists have, therefore, long worked to create software that can do the job automatically. DeepMind was the first to accomplish that goal three years ago with its AlphaFold model, which showed that it could predict the structure of proteins within a few days.
The Google unit recently unveiled a new version of AlphaFold that has improved prediction capabilities. DeepMind claims that it can estimate the shapes of various biological molecules in addition to proteins.
Among other things, the improved version of AlphaFold is useful for ligand structure prediction. These are molecules that have the ability to attach to proteins and alter how those proteins function. Ligands are essential to cell signaling, a vital biological mechanism by which cells affect one another’s behavior.
It is also possible for AlphaFold to estimate the structure of other molecules. Nucleic acids, a class of compounds that includes DNA and RNA, are among those molecules. Furthermore, according to DeepMind, AlphaFold can now calculate the shapes of more molecules with greater accuracy.
A protein-ligand complex results from a ligand attaching or binding to a protein. Historically, researchers used a docking technique to assess these complexes’ shape. This approach is only practical when a substantial body of data regarding the protein component of a protein-ligand complex is accessible.
DeepMind claims that the most recent iteration of AlphaFold can more precisely predict the structure of protein-ligand complexes than the best docking techniques. Moreover, it accomplishes this with a far smaller amount of data needed than those approaches. Therefore, AlphaFold may facilitate researchers’ ability to examine recently identified protein-ligand complexes for which there is scant data.
According to DeepMind, the AI model also offers improved accuracy in other domains. The Google unit claims that AlphaFold is able to predict the structure of almost every molecule in Protein Data Bank, a popular scientific database. According to DeepMind, the model frequently produces those forecasts with “atomic accuracy.”
DeepMind researchers wrote in a blog post, “Early analysis also shows that our model greatly outperforms AlphaFold2.3 on some protein structure prediction problems that are relevant for drug discovery, like antibody binding. Additionally, accurately predicting protein-ligand structures is an incredibly valuable tool for drug discovery, as it can help scientists identify and design new molecules, which could become drugs.”
AlphaFold is already being used by scientists to assist with research initiatives. More than 1.4 million people have accessed the AlphaFold Protein Structure Database, which houses protein structures produced by the AI model, according to a recent disclosure from DeepMind. Moreover, AlphaFold has been integrated by DeepMind spinoff Isomorphic Labs into its drug discovery endeavors.