Tackling APOE's structural challenges via in silico modeling in the era of neural networks: Can AlphaFold II help circumvent the problem of lacking full-length protein structure?

Abstract

The APOE gene, encoding apolipoprotein E, is a predictor of longevity and age-related diseases. Despite numerous genetic studies, the data on molecular mechanisms by which apolipoprotein E affects the human phenotype remain incomplete due to the structural properties of the protein. Recently, a number of studies have used in silico drug discovery techniques based on protein-ligand docking, further highlighting the issue of lacking 3D structure of apolipoprotein E. Using molecular dynamics simulation, we found that AlphaFold II models of apolipoprotein E conformationally significantly differ both from the only available NMR structure, 2L7B, and structures obtained through circular dichroism spectroscopy: the ε4 isoform lacks the salt bridge between R61 and E255, while the ε2 and ε3 isoforms have extensive networks of interdomain interactions. Our findings challenge the benefits of using AlphaFold II for obtaining starting conformations for molecular docking.