Search for: All records

Variational autoencoders are artificial neural networks with the capability to reduce highly dimensional sets of data to smaller dimensional, latent representations. In this work, these models are applied to molecular dynamics simulations of the self-assembly of coarse-grained peptides to obtain a singled-valued order parameter for amyloid aggregation. This automatically learned order parameter is constructed by time-averaging the latent parameterizations of internal coordinate representations and compared to the nematic order parameter which is commonly used to study ordering of similar systems in literature. It is found that the latent space value provides more tailored insight into the aggregation mechanism’s details, correctly identifying fibril formation in instances where the nematic order parameter fails to do so. A means is provided by which the latent space value can be analyzed so that the major contributing internal coordinates are identified, allowing for a direct interpretation of the latent space order parameter in terms of the behavior of the system. The latent model is found to be an effective and convenient way of representing the data from the dynamic ensemble and provides a means of reducing the dimensionality of a system whose scale exceeds molecular systems so-far considered with similar tools. This bypasses a need for researcher speculation on what elements of a system best contribute to summarizing major transitions, and suggests latent models are effective and insightful when applied to large systems with a diversity of complex behavior. 

Proteins and peptides in nature are almost exclusively made from l -amino acids, and this is even more absolute in the metazoan. With the advent of modern bioanalytical techniques, however, previously unappreciated roles for d -amino acids in biological processes have been revealed. Over 30 d -amino acid containing peptides (DAACPs) have been discovered in animals where at least one l -residue has been isomerized to the d -form via an enzyme-catalyzed process. In Aplysia californica , GdFFD and GdYFD (the lower-case letter “d” indicates a d -amino acid residue) modulate the feeding behavior by activating the Aplysia achatin-like neuropeptide receptor (apALNR). However, little is known about how the three-dimensional conformation of DAACPs influences activity at the receptor, and the role that d -residues play in these peptide conformations. Here, we use a combination of computational modeling, drift-tube ion-mobility mass spectrometry, and receptor activation assays to create a simple model that predicts bioactivities for a series of GdFFD analogs. Our results suggest that the active conformations of GdFFD and GdYFD are similar to their lowest energy conformations in solution. Our model helps connect the predicted structures of GdFFD analogs to their activities, and highlights a steric effect on peptide activity at position 1 on the GdFFD receptor apALNR. Overall, these methods allow us to understand ligand–receptor interactions in the absence of high-resolution structural data.