The occurrence of linear cholesterol-recognition motifs in alpha-helical transmembrane domains has long been debated. Here, we demonstrate the ability of a genetic algorithm guided by coarse-grained molecular dynamics simulations---a method coined evolutionary molecular dynamics (evo-MD)---to directly resolve the sequence which maximally attracts/sorts cholesterol within a single-pass alpha-helical transmembrane domain (TMDs). We illustrate that the evolutionary landscape of cholesterol attraction in membrane proteins is characterized by a sharp, well-defined global optimum. Surprisingly, this optimal solution features an unusual short hydrophobic block, consisting of typically only eight short chain hydrophobic amino acids, surrounded by three successive lysines. Owing to the membrane thickening effect of cholesterol, cholesterol-enriched ordered phases favor TMDs characterized by a long rather than a short hydrophobic length. However, this short hydrophobic pattern evidently offers a pronounced net advantage for the binding of free cholesterol in both coarse-grained and atomistic simulations. Attraction is mediated by the unique ability of cholesterol to snorkel within the hydrophobic core of the membrane and thereby shield deeply located lysines from the unfavorable hydrophobic surrounding. Since this mechanism of attraction is of a thermodynamic nature and is not based on molecular shape specificity, a large diversity of sub-optimal cholesterol attracting sequences can exist. The puzzling sequence variability of proposed linear cholesterol-recognition motifs is thus consistent with sub-optimal, unspecific binding of cholesterol. Importantly, since evo-MD uniquely enables the targeted design of recognition motifs for distinct fluid lipid membranes, we foresee wide applications for evo-MD in the biological and biomedical fields.

Solid substrates often induce non-uniform strain and doping in graphene monolayer, therefore altering the intrinsic properties of graphene, reducing its charge carrier mobilities and, consequently, the overall electrical performance. Here, we exploit confocal Raman spectroscopy to study graphene directly free-floating on the surface of water, and show that liquid supports relief the preexisting strain, have negligible doping effect and restore the uniformity of the properties throughout the graphene sheet. Such an effect originates from the structural adaptability and flexibility, lesser contamination and weaker intermolecular bonding of liquids compared to solid supports, independently of the chemical nature of the liquid. Moreover, we demonstrate that water provides a platform to study and distinguish chemical defects from substrate-induced defects, in the particular case of hydrogenated graphene. Liquid supports, thus, are advantageous over solid supports for a range of applications, particularly for monitoring changes in the graphene structure upon chemical modification.

Proteins can specifically bind to curved membranes through curvature-induced hydrophobic lipid packing defects. The chemical diversity among such curvature “sensors” challenges our understanding of how they differ from general membrane “binders” that bind without curvature selectivity. Here, we combine an evolutionary algorithm with coarse-grained molecular dynamics simulations (Evo-MD) to resolve the peptide sequences that optimally recognize the curvature of lipid membranes. We subsequently demonstrate how a synergy between Evo-MD and a neural network (NN) can enhance the identification and discovery of curvature sensing peptides and proteins. To this aim, we benchmark a physics-trained NN model against experimental data and show that we can correctly identify known sensors and binders. We illustrate that sensing and binding are phenomena that lie on the same thermodynamic continuum, with only subtle but explainable differences in membrane binding free energy, consistent with the serendipitous discovery of sensors.

Proteins can sense hydrophobic lipid packing defects on positively curved membranes. The chemical diversity mong such 'sensors' challenges our understanding of how they differ from 'binders', that bind membranes without curvature selectivity. Here, we combine an evolutionary algorithm with coarse-grained molecular dynamics (Evo-MD) to inverse design curvature sensing peptides. The resolved optimum illustrates that curvature sensing is driven by hydrophobic interactions. With the data obtained in the design process, we trained a neural network (NN) that can predict sensing free energy from a peptide sequence only. We applied this NN to experimentally classified sensors and binders and illustrate why sensing and binding are phenomena on the same thermodynamic continuum, with only subtle differences in membrane binding free energy. The here-developed physics-based inverse design approach introduces a quantum leap in the computational design of peptide drugs that recognise (inter)facial characteristics of membranes in viruses, bacteria and cancer cells.

MotivationMany membrane peripheral proteins have evolved to transiently interact with the surface of (curved) lipid bilayers. Currently, methods toquantitativelypredict sensing and binding free energies for protein sequences or structures are lacking, and such tools could greatly benefit the discovery of membrane-interacting motifs, as well as theirde novodesign.ResultsHere, we trained a transformer neural network model on molecular dynamics data for>50,000 peptides that is able to accurately predict the (relative) membrane-binding free energy for any given amino acid sequence. Using this information, our physics-informed model is able to classify a peptide’s membrane-associative activity as either non-binding, curvature sensing, or membrane binding. Moreover, this method can be applied to detect membrane-interaction regions in a wide variety of proteins, with comparable predictive performance as state-of-the-art data-driven tools like DREAMM, PPM, and MODA, but with a wider applicability regarding protein diversity, and the added feature to distinguish curvature sensing from general membrane binding.AvailabilityWe made these tools available as a web server, coined Protein-Membrane Interaction predictor (PMIpred), which can be accessed athttps://pmipred.fkt.physik.tu-dortmund.de.