Share this story!
- The AI AlphaFold expands beyond proteins to various biomolecule structures.
- This opens new horizons in drug discovery and disease understanding.
- A step towards digital biology through enhanced molecular modeling.
AlphaFold, by Google DeepMind, initially aimed at decoding the mysteries of protein folding, has broadened its molecular understanding beyond proteins to include ligands, nucleic acids (DNA and RNA), and post-translational modifications.
This evolution opens a realm of practical applications, notably accelerating drug discovery, which could fast-track the path to novel treatments for various diseases.
A leap beyond protein folding
AlphaFold commenced its journey focusing on single-chain protein prediction, a cornerstone for understanding biological processes. Its evolution over time, marked by the development of AlphaFold-Multimer and AlphaFold 2.3, saw an expanded coverage to larger complexes and nearly all cataloged proteins known to science.
In 2022, the model's structure predictions were made freely available through the AlphaFold Protein Structure Database, in partnership with EMBL's European Bioinformatics Institute (EMBL-EBI).
This open-access database has since been instrumental for researchers worldwide, aiding advancements in areas like malaria vaccines, cancer drug discovery, and the development of plastic-eating enzymes to combat pollution.
Speeding up drug discovery
Early analysis shows the model's superior performance in some protein structure prediction problems relevant to drug discovery, like antibody binding.
The accurate prediction of protein-ligand structures, a critical aspect of drug discovery, has been significantly improved.
Unlike conventional 'docking methods', which require a rigid reference protein structure, the latest AlphaFold model sets a new standard by accurately predicting structures without needing a reference, allowing for predictions of completely novel proteins.
AlphaFold's ability to predict protein-ligand structures is crucial for therapeutic drug design. This is illustrated by its accurate predictions in several cases. For example, it accurately predicted structures of PORCN, a clinical-stage anti-cancer molecule. Similarly, it did well with KRAS, a complex with a covalent ligand targeting cancer. It also accurately modeled PI5P4Kγ, a selective inhibitor of a lipid kinase relevant for cancer and immunological disorders. Each case shows AlphaFold's practical utility in drug design.
Better understanding of fundamental biology
The model also facilitates a better understanding of fundamental biology. A notable instance is the structure prediction of CasLambda bound to crRNA and DNA, part of the CRISPR family, known for its genome editing capability. This insight not only augments our understanding of genetic mechanisms but also expedites therapeutic applications of such knowledge.
No AlphaFold with an AI pause
In the spring several prominent people proposed a pause on AI development. If that pause had been enacted, this version of AlphaFold would not have seen the light of day now.
Harsh regulations or "pauses" would seriously harm AI development, and slow important processes. Faster and better drug discovery will save lives. An AI pause is bad for your health.