Physics-based generative model of curvature sensing peptides; distinguishing sensors from binders

Abstract

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.

PMIpred: A physics-informed web server for quantitative Protein-Membrane Interaction prediction

“…Details on the derivations of eq. 1-4 and their constants can be found in our previous publication (van Hilten et al [2023]).…”

“… A) The membrane-binding probability P m as a function of the membrane-binding free energy Δ F sm (at R = 50 nm) shows a sharp transition. The orange area marks the ‘sensor regime’ (van Hilten et al [2023]). Insets show cartoon explanations of the three classes.…”

“…MD-calculated values are averages and standard deviations for 3 independent replicas (500 ns simulation per run, as described in ref. van Hilten et al [2023]). C) Schematic workflow of predicting membrane-interaction regions in protein structures within PMIpred.…”

“…Our approach involves a transformer neural network (NN) model that is trained on a large set of molecular dynamics (MD) data on relative binding free energies for short peptides (24 residues) interacting with model membranes. Since our data were initially gathered during the optimisation of curvature sensing (van Hilten et al [2023]), our model is additionally able to distinguish curvature- (or, equivalently, lipid packing defect-) sensing from general membrane binding. Moreover, because of its physics-informed character, it is – compared to previous data-driven methods – better able to generalise membrane-interaction features across a wide variety of protein families, including lipid kinases, ALPS-containing curvature sensors, α -synuclein, and N-BAR proteins.…”

Physics-based inverse design of cholesterol attracting transmembrane helices reveals a paradoxical role of hydrophobic length

“…We therefore expect Evo-MD to yield novel insights into the mechanisms that underpin the molecular organization of biological membranes and protein trafficking. In particular, we have recently applied Evo-MD to the design of peptide motifs capable of selectively targeting membrane curvature (49), and are currently exploring the targeting of other characteristic membrane features such as lipid composition. Importantly, this will pave the road for the inverse design of peptide drugs and peptide based drug vehicles capable of selectively targeting the fluid membranes of viruses, microbes, and cancer cells; since their membrane leaflets are characterized by pronounced differences in curvature (50) and/or lipid composition (51).…”