Research
Progress update: Our latest AlphaFold model shows significantly improved accuracy and extends coverage beyond proteins to other biological molecules, including ligands
Since its launch in 2020, AlphaFold has revolutionized the way proteins and their interactions are understood. Google DeepMind and Isomorphic Labs have collaborated to build the foundation of a more powerful AI model that extends coverage beyond proteins to the full range of biologically relevant molecules.
Today we are notification of an update on progress towards the next generation of AlphaFold. Our latest model can now generate predictions for almost all of its molecules Protein Data Bank (PDB), often reaching atomic precision.
It unlocks new understanding and greatly improves accuracy across multiple key classes of biomolecules, including ligands (small molecules), proteins, nucleic acids (DNA and RNA), and those containing post-translational modifications (PTMs). These different types of structure and complexes are essential for understanding the biological mechanisms inside the cell and have been difficult to predict with high accuracy.
The model’s expanded capabilities and performance can help accelerate biomedical discoveries and realize the next era of “digital biology” — yielding new insights into the function of disease pathways, genomics, biorenewable materials, plant immunity , potential therapeutic targets, drug design mechanisms, and novel platforms for enabling protein engineering and synthetic biology.
Series of predicted structures compared to the ground truth (white) from our latest AlphaFold model.
Above and beyond the protein fold
AlphaFold was a seminal breakthrough for single-chain protein prediction. AlphaFold-Multimer then extended to complexes with multiple protein chains, followed by AlphaFold2.3, which improved performance and extended coverage to larger complexes.
In 2022, AlphaFold’s structure predictions for nearly all proteins known to science were made freely available through the AlphaFold Protein Structure Databasein collaboration with EMBL’s European Bioinformatics Institute (EMBL-EBI).
To date, 1.4 million users in more than 190 countries have accessed the AlphaFold database, and scientists around the world have used AlphaFold’s predictions to help advance research into everything from accelerating new malaria vaccines to and advancing the discovery of anti-cancer drugs to the development of plastic-consuming enzymes to tackle pollution. .
Here we show the remarkable abilities of AlphaFold to predict accurate structures beyond protein folding, generating highly accurate structure predictions in ligands, proteins, nucleic acids, and post-translational modifications.
Performance between protein-ligand complexes (a), proteins (b), nucleic acids (c) and covalent modifications (d).
Accelerated drug discovery
Early analysis also shows that our model far outperforms AlphaFold2.3 in some protein structure prediction problems relevant to drug discovery, such as 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 that could become drugs.
The current industry standard is to use “docking methods” to determine interactions between ligands and proteins. These docking methods require a rigid reference protein structure and a proposed site for ligand binding.
Our latest model sets a new bar for protein-ligand structure prediction by outperforming the best reported docking methods, without requiring a reference protein structure or the location of the ligand pocket—allowing predictions of entirely new proteins that have not been structurally characterized before .
It can also model the positions of all atoms together, allowing it to represent the full inherent flexibility of proteins and nucleic acids as they interact with other molecules—something not possible using docking methods.
Here, for example, are three recently published, therapeutically relevant cases where the predicted structures of our latest model (shown in color) closely match the experimentally determined structures (shown in gray):
- PORKN: A clinical-stage anticancer molecule bound to its target, along with another protein.
- KRAS: Ternary complex with a covalent linker (molecular glue) of an important cancer target.
- PI5P4Kc: Selective allosteric inhibitor of a lipid kinase, with multiple disease implications, including cancer and immune disorders.
Predictions for PORCN (1), KRAS (2) and PI5P4Kγ (3).
Isomorphic Labs applies this next-generation AlphaFold model to therapeutic drug design, helping to rapidly and accurately characterize many types of macromolecular structures relevant to disease treatment.
New understanding of biology
Unlocking the modeling of protein and ligand structures along with nucleic acids and those containing post-translational modifications, our model provides a faster and more accurate tool for examining fundamental biology.
An example includes its structure CasLambda binds to crRNA and DNAPart of the CRISPR family. CasLambda shares its genome editing capability CRISPR-Cas9 system, commonly known as “genetic scissors,” which researchers can use to alter the DNA of animals, plants, and microorganisms. The smaller size of CasLambda may allow more efficient use in genome editing.
Predicted structure of crRNA- and DNA-binding CasLambda (Cas12l), part of the CRISPR subsystem.
The latest version of AlphaFold’s ability to model such complex systems shows us that artificial intelligence can help us better understand these types of mechanisms and accelerate their use for therapeutic applications. More examples are available in our progress update.
Advancing scientific exploration
The dramatic jump in our model’s performance shows the potential for artificial intelligence to greatly enhance scientific understanding of the molecular machines that make up the human body — and the wider natural world.
AlphaFold has already catalyzed important scientific advances around the world. Now, the next generation of AlphaFold has the potential to help advance scientific exploration at digital speed.
Our dedicated teams at Google DeepMind and Isomorphic Labs have made great strides forward in this critical work, and we look forward to sharing our continued progress.