– DeepMind announced AlphaFold 3, the latest version of its protein folding project for predicting how proteins fold based on amino acid sequences
– Proteins fold like ‘origami’ into 3D structures that determine their functions, and misfolding can contribute to diseases like Alzheimer’s and Parkinson’s
– AlphaFold 3 uses deep learning to predict protein structures and has evolved to incorporate DNA, RNA, and small molecules interactions, with applications in drug design and a user-friendly AlphaFold Server platform for researchers.
DeepMind recently introduced AlphaFold 3, the latest version of its protein folding project, which predicts how proteins fold based on their amino acid sequences. Proteins fold into 3D structures like ‘origami,’ impacting their functions. Understanding protein folding is crucial in biochemistry and molecular biology as proteins are essential for life and can become misfolded, leading to diseases like Alzheimer’s and Parkinson’s.
AlphaFold uses machine learning to simulate protein folding, addressing the challenge of predicting the correct structure due to the high number of possible configurations. AlphaFold 3 features an improved Evoformer module and a diffusion network to assemble predicted structures accurately. The model encompasses not only proteins but also DNA, RNA, and small molecules, enhancing its applications in various fields.
The AlphaFold project began in 2016 and has evolved with versions like AlphaFold 1 and AlphaFold 2, which achieved high accuracy in protein structure predictions. The latest iteration, AlphaFold 3, continues to push boundaries in protein discovery and analysis. The AlphaFold Protein Structure Database, released in partnership with EMBL-EBI, provides access to over 350,000 protein structure predictions, benefiting researchers in various fields worldwide. With over one million users in 190 countries, the database has facilitated discoveries across medicine, agriculture, and beyond.