The Dependence of Amyloid-β Dynamics on Protein Force Fields and Water Models

We studied the dynamics of A(40), involved in Alzheimer's disease, by using 21 methods combined from Amber03, Amber99sb-ILDN, Charmm27, Charmm22*, OPLS-2001, OPLS-2006, OPLS-2008, Gromos96-43a1, Gromos96-53a6, Gromos96-54a7, and the water models SPC, TIP3P, TIP4P. Major differences in the structural ensembles were systematized: Amber03, Charmm27, and Gromos96-54a7 stabilize the helices; Gromos96-43a1 and Gromos53a6 favor the -strands (with Charmm22* and Amber99sb-ILDN in between), and OPLS produces unstructured ensembles. The accuracy of the NMR chemical shifts was in the order: Charmm22*>Amber99sb-ILDN>OPLS-2008 approximate to Gromos96-43a1>Gromos96-54a7 approximate to OPLS-2001>OPLS-2006>Gromos96-53a6>Charmm27>Amber03. The computed (3)J(HNH)-coupling constants were sensitive to experiment type and Karplus parameterization. Overall, the ensembles of Charmm22* and Amber99sb-ILDN provided the best agreement with experimental NMR and circular dichroism data, providing a model for the real A monomer ensemble. Also, the polar water model TIP3P significantly favored helix and compact conformations.