AlphaFold-multimer predicts ATG8 protein binding motifs crucial for autophagy research

Abstract

Proteins are structural and executing macromolecules essential for life in all biological systems. Insight into proteins structures is required for detailed mechanistic understanding of how they work and solve different tasks. The ability to predict three-dimensional (3D) protein structures from primary sequence information has therefore been an open research question for more than 50 years. The search for this holy grail of structural bioinformatics has recently led to development of AlphaFold2, an amazing artificial intelligence–based structure prediction tool, developed by scientists from Google DeepMind [1]. How does AlphaFold2 work? Very briefly, it searches sequence databases to find sequences similar to the input, produces a multiple sequence alignment, uses a neural network to extract information, and sends that information to a second neural network that calculates a 3D structure. This is done in an iterative manner. In this issue of PLOS Biology, Ibrahim and colleagues demonstrate how AlphaFold2 (AF2)-multimer can be used as an important and powerful new tool to successfully predict so-called LC3 interacting region (LIR) motifs (see below) in proteins involved in autophagy processes [2].