Search

Our method identifies outliers in a latent space model of the system's sampled conformations that is periodically trained using a convolutional variational autoencoder. As a proof of principle, we applied our DL-enhanced WE method to simulate a millisecond protein folding process. To enable rapid tests, our simulations propagated discrete-state synthetic molecular dynamics trajectories using a generative, fine-grained Markov state model. Results revealed that our ″on-the-fly″ DL of outliers enhanced the efficiency of WE by >3-fold in estimating the folding rate constant. Our efforts are a significant step forward in the unsupervised learning of slow coordinates during rare event sampling.

Despite their widespread use as therapeutics, clinical development of small molecule drugs remains challenging. Among the many parameters that undergo optimization during the drug development process, increasing passive cell permeability (i.e., log(P)) can have some of the largest impact on potency. Cyclic peptides (CPs) have emerged as a viable alternative to small molecules, as they retain many of the advantages of small molecules (oral availability, target specificity) while being highly effective at traversing the plasma membrane. However, the relationship between the dominant conformations that typify CPs in an aqueous versus a membrane environment and cell permeability remain poorly characterized. In this study, we have used Gaussian accelerated molecular dynamics (GaMD) simulations to characterize the effect of solvent on the free energy landscape of lariat peptides, a subset of CPs that have recently shown potential for drug development (Kelly et al., JACS 2021). Differences in the free energy of lariat peptides as a function of solvent can be used to predict permeability of these molecules, and our results show that permeability is most greatly influenced by N-methylation and exposure to solvent. Our approach lays the groundwork for using GaMD as a way to virtually screen large libraries of CPs and drive forward development of CP-based therapeutics., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Frazee et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Huntington's disease is a neurodegenerative disorder caused by an expanded polyglutamine (polyQ) domain within the huntingtin protein (htt) that initiates toxic protein aggregation. Htt directly interacts with membranes, influencing aggregation and spurring membrane abnormalities. These interactions are facilitated by the 17 N-terminal residues (Nt17) that form an amphipathic α-helix implicated in both lipid binding and aggregation. Here, the impact of unsaturation in phospholipid tails on htt-lipid interaction and htt aggregation was determined. There was no correlation between the degree of htt-lipid complexation and the degree of htt aggregation in the presence of each lipid system, indicating that lipid systems with different properties uniquely alter the membrane-mediated aggregation mechanisms. Also, the association between Nt17 and membrane surfaces is determined by complementarity between hydrophobic residues and membrane defects and how easily the peptide can partition into the bilayer. Our results provide critical insights into how membrane physical properties influence downstream htt aggregation.

The pH-Low Insertion Peptide (pHLIP) is a membrane-active peptide that spontaneously folds into a transmembrane α-helix upon acidification. This activity enables pHLIP to potentially act as a vector for drugs related to diseases characterized by acidosis such as cancer or heart ischemia. Presently, due to aggregation-based effects, formulations of pHLIP are only viable at near-μM concentrations. In addition, since most of pHLIP's measurable qualities involve a membrane, probing the details of pHLIP in the interstitial region is difficult. In attempts to shed light on these issues, we performed constant pH molecular dynamics simulations on pHLIP as well as P20G, a variant with increased helicity, in solution at 0 and 150 mM NaCl over a broad range of pHs. In general, the addition of ions reduced the effective pK a of the acidic residues in pHLIP. P20G exhibits a higher helicity than pHLIP in general and is more compact than pHLIP at pH values under 4. In terms of charge effects, sodium cations localized predominantly to the C-terminus of the peptide with a high density of acidic residues. Additionally, the salt bridge between R11 and D14 is by far the most favored and particularly so with pHLIP at 150 mM NaCl. We expect that this approach will be a valuable tool to screen variants of pHLIP for favorable properties in solution, an aspect of pHLIP design that to this point has largely been neglected., (© 2021 Wiley Periodicals LLC.)

Membrane-active peptides (MAPs) are short-length peptides used for potential biomedical applications in diagnostic imaging of tissues, targeted drug delivery, gene delivery, and antimicrobials and antibiotics. The broad appeal of MAPs is that they are infinitely variable, relatively low cost, and biocompatible. However, experimentally characterizing the specific properties of a MAP or its many variants is a low-resolution and potentially time-consuming endeavor; molecular dynamics (MD) simulations have emerged as an invaluable tool in identifying the biophysical interactions that are fundamental to the function of MAPs. In this chapter, a step-by-step approach to discreetly model the binding, folding, and insertion of a membrane-active peptide to a model lipid bilayer using MD simulations is described. Detailed discussion is devoted to the critical aspects of running these types of simulations: prior knowledge of the system, understanding the strengths and weaknesses of molecular mechanics force fields, proper construction and equilibration of the system, realistically estimating both experimental and computational timescales, and leveraging analysis to make direct comparisons to experimental results as often as possible., (© 2021. Springer Science+Business Media, LLC, part of Springer Nature.)