UC Santa Barbara

Amyloid fibrils are a solid composed of proteins or peptides arranged in what is known as a cross beta pattern. That is, the fibril is made of beta sheets where the peptide backbone is perpendicular to the fibril axis. This structure has been associated with several difficult to treat diseases including Alzheimer’s Disease, type II diabetes, and amyotrophic lateral sclerosis. The characterization of the formation of these fibrils remains poorly understood at a fundamental level.

First, I consider the understanding of the primary structure – activity relationship for amyloid-forming peptides. Here I use a neural-network based method of analysis: the classifying autoencoder (CAE). I demonstrate its capabilities by applying the technique to an experimental database (the Waltz database) to provide insight into a novel descriptor, dimeric isotropic deviation — an experimental measure of the aggregation properties of the amino acids. I find correlation between dimeric isotropic deviation and the failure to form amyloids when hydrophobic effects are not a primary driving force in amyloid formation.

Next, I consider the formation of amyloids from the perspective of a molecular dynamics simulation. I use a similar technique as above to analyze molecular dynamics simulations of amyloid formation. Here we the technique is applied to the internal coordinates of a coarse-grained molecular dynamics simulation of amyloid formation. The method is shown to be able to reduce the ensemble of data to a single variable that tracks evolution in the system and successfully characterizes large-scale system

evolutions with precision exceeding or comparable to more conventional order parameters. In addition, we show it can be used to identify the features of the system which best track the evolution of the system and be used to automatically detect the nucleus of the aggregating system.