Published in:Journal of chemical theory and computation 16(12), 7825 - 7839 (2020). doi:10.1021/acs.jctc.0c00727
Language:
English
DOI:
https://doi.org/10.1021/acs.jctc.0c00727
ISSN:
1549-9618;
1549-9626
Origination:
Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
Molecular dynamic (MD) simulations are animportant tool for studying protein aggregation processes, whichplay a central role in a number of diseases including Alzheimer’sdisease. However, MD simulations produce large amounts of data,requiring advanced methods to extract mechanistic insight into theprocess under study. Transition networks (TNs) provide anelegant method to identify (meta)stable states and the transitionsbetween them from MD simulations. Here, we apply two differentmethods to generate TNs for protein aggregation: Markov statemodels (MSMs), which are based on kinetic clustering the statespace, and TNs using conformational clustering. The similaritiesand differences of both methods are elucidated for the aggregationof the fragment Aβ16−22 of the Alzheimer’s amyloid-β peptide. Ingeneral, both methods perform excellently in identifying the main aggregation pathways. The strength of MSMs is that they providea rather coarse and thus simply to interpret picture of the aggregation process. Conformation-sorting TNs, on the other hand,outperform MSMs in uncovering mechanistic details. We thus recommend to apply both methods to MD data of proteinaggregation in order to obtain a complete picture of this process. As part of this work, a Python script called ATRANET forautomated TN generation based on a correlation analysis of the descriptors used for conformational sorting is made publiclyavailable.