Highly accurate protein structure prediction with AlphaFold
Presents AlphaFold, a deep learning system that predicts atomic-accuracy protein 3D structures from amino acid sequence alone.
Predicting a protein's 3D structure from its amino acid sequence has been an open problem for over 50 years; existing methods fell short of atomic accuracy, especially without homologous structures. This paper presents a redesigned AlphaFold, validated in CASP14, as the first method to regularly predict atomic-accuracy structures even without similar known structures, competitive with experimental accuracy in most cases, using a deep learning architecture that incorporates physical and biological knowledge, including multi-sequence alignments.
Based on: Highly accurate protein structure prediction with AlphaFold · Nature
Curated by Aramai Editorial
Read summary →