When Data Are Lacking: Physics-Based Inverse Design of Biopolymers Interacting with Complex, Fluid Phases.

Biomolecular research traditionally revolves around comprehending the mechanisms through which peptides or proteins facilitate specific functions, often driven by their relevance to clinical ailments. This conventional approach assumes that unraveling mechanisms is a prerequisite for wielding control over functionality, which stands as the ultimate research goal. However, an alternative perspective emerges from physics-based inverse design, shifting the focus from mechanisms to the direct acquisition of functional control strategies. By embracing this methodology, we can uncover solutions that might not have direct parallels in natural systems, yet yield crucial insights into the isolated molecular elements dictating functionality. This provides a distinctive comprehension of the underlying mechanisms.In this context, we elucidate how physics-based inverse design, facilitated by evolutionary algorithms and coarse-grained molecular simulations, charts a promising course for innovating the reverse engineering of biopolymers interacting with intricate fluid phases such as lipid membranes and liquid protein phases. We introduce evolutionary molecular dynamics (Evo-MD) simulations, an approach that merges evolutionary algorithms with the Martini coarse-grained force field. This method directs the evolutionary process from random amino acid sequences toward peptides interacting with complex fluid phases such as biological lipid membranes, offering significant promises in the development of peptide-based sensors and drugs. This approach can be tailored to recognize or selectively target specific attributes such as membrane curvature, lipid composition, membrane phase (e.g., lipid rafts), and protein fluid phases. Although the resulting optimal solutions may not perfectly align with biological norms, physics-based inverse design excels at isolating relevant physicochemical principles and thermodynamic driving forces governing optimal biopolymer interaction within complex fluidic environments. In addition, we expound upon how physics-based evolution using the Evo-MD approach can be harnessed to extract the evolutionary optimization fingerprints of protein-lipid interactions from native proteins. Finally, we outline how such an approach is uniquely able to generate strategic training data for predictive neural network models that cover the whole relevant physicochemical domain. Exploring challenges, we address key considerations such as choosing a fitting fitness function to delineate the desired functionality. Additionally, we scrutinize assumptions tied to system setup, the targeted protein structure, and limitations posed by the utilized (coarse-grained) force fields and explore potential strategies for guiding evolution with limited experimental data. This discourse encapsulates the potential and remaining obstacles of physics-based inverse design, paving the way for an exciting frontier in biomolecular research.

CGCompiler: Automated Coarse-Grained Molecule Parametrization via Noise-Resistant Mixed-Variable Optimization.

Coarse-grained force fields (CG FFs) such as the Martini model entail a predefined, fixed set of Lennard-Jones parameters (building blocks) to model virtually all possible nonbonded interactions between chemically relevant molecules. Owing to its universality and transferability, the building-block coarse-grained approach has gained tremendous popularity over the past decade. The parametrization of molecules can be highly complex and often involves the selection and fine-tuning of a large number of parameters (e.g., bead types and bond lengths) to optimally match multiple relevant targets simultaneously. The parametrization of a molecule within the building-block CG approach is a mixed-variable optimization problem: the nonbonded interactions are discrete variables, whereas the bonded interactions are continuous variables. Here, we pioneer the utility of mixed-variable particle swarm optimization in automatically parametrizing molecules within the Martini 3 coarse-grained force field by matching both structural (e.g., RDFs) as well as thermodynamic data (phase-transition temperatures). For the sake of demonstration, we parametrize the linker of the lipid sphingomyelin. The important advantage of our approach is that both bonded and nonbonded interactions are simultaneously optimized while conserving the search efficiency of vector guided particle swarm optimization (PSO) methods over other metaheuristic search methods such as genetic algorithms. In addition, we explore noise-mitigation strategies in matching the phase-transition temperatures of lipid membranes, where nucleation and concomitant hysteresis introduce a dominant noise term within the objective function. We propose that noise-resistant mixed-variable PSO methods can both improve and automate parametrization of molecules within building-block CG FFs, such as Martini.

TatA and TatB generate a hydrophobic mismatch important for the function and assembly of the Tat translocon in Escherichia coli.

The twin-arginine translocation (Tat) system serves to translocate folded proteins across energy-transducing membranes in bacteria, archaea, plastids, and some mitochondria. In Escherichia coli, TatA, TatB, and TatC constitute functional translocons. TatA and TatB both possess an N-terminal transmembrane helix (TMH) followed by an amphipathic helix. The TMHs of TatA and TatB generate a hydrophobic mismatch with the membrane, as the helices comprise only 12 consecutive hydrophobic residues; however, the purpose of this mismatch is unclear. Here, we shortened or extended this stretch of hydrophobic residues in either TatA, TatB, or both and analyzed effects on translocon function and assembly. We found the WT length helices functioned best, but some variation was clearly tolerated. Defects in function were exacerbated by simultaneous mutations in TatA and TatB, indicating partial compensation of mutations in each by the other. Furthermore, length variation in TatB destabilized TatBC-containing complexes, revealing that the 12-residue-length is important but not essential for this interaction and translocon assembly. To also address potential effects of helix length on TatA interactions, we characterized these interactions by molecular dynamics simulations, after having characterized the TatA assemblies by metal-tagging transmission electron microscopy. In these simulations, we found that interacting short TMHs of larger TatA assemblies were thinning the membrane and-together with laterally-aligned tilted amphipathic helices-generated a deep V-shaped membrane groove. We propose the 12 consecutive hydrophobic residues may thus serve to destabilize the membrane during Tat transport, and their conservation could represent a delicate compromise between functionality and minimization of proton leakage.

Efficient Quantification of Lipid Packing Defect Sensing by Amphipathic Peptides: Comparing Martini 2 and 3 with CHARMM36.

In biological systems, proteins can be attracted to curved or stretched regions of lipid bilayers by sensing hydrophobic defects in the lipid packing on the membrane surface. Here, we present an efficient end-state free energy calculation method to quantify such sensing in molecular dynamics simulations. We illustrate that lipid packing defect sensing can be defined as the difference in mechanical work required to stretch a membrane with and without a peptide bound to the surface. We also demonstrate that a peptide's ability to concurrently induce excess leaflet area (tension) and elastic softening─a property we call the "characteristic area of sensing" (CHAOS)─and lipid packing sensing behavior are in fact two sides of the same coin. In essence, defect sensing displays a peptide's propensity to generate tension. The here-proposed mechanical pathway is equally accurate yet, computationally, about 40 times less costly than the commonly used alchemical pathway (thermodynamic integration), allowing for more feasible free energy calculations in atomistic simulations. This enabled us to directly compare the Martini 2 and 3 coarse-grained and the CHARMM36 atomistic force fields in terms of relative binding free energies for six representative peptides including the curvature sensor ALPS and two antiviral amphipathic helices (AH). We observed that Martini 3 qualitatively reproduces experimental trends while producing substantially lower (relative) binding free energies and shallower membrane insertion depths compared to atomistic simulations. In contrast, Martini 2 tends to overestimate (relative) binding free energies. Finally, we offer a glimpse into how our end-state-based free energy method can enable the inverse design of optimal lipid packing defect sensing peptides when used in conjunction with our recently developed evolutionary molecular dynamics (Evo-MD) method. We argue that these optimized defect sensors─aside from their biomedical and biophysical relevance─can provide valuable targets for the development of lipid force fields.

Quantifying Membrane Curvature Sensing of Peripheral Proteins by Simulated Buckling and Umbrella Sampling.

Membrane curvature plays an essential role in the organization and trafficking of membrane associated proteins. Comparison or prediction of the experimentally resolved protein concentrations adopted at different membrane curvatures requires direct quantification of the relative partitioning free energy. Here, we present a highly efficient and simple to implement a free-energy calculation method which is able to directly resolve the relative partitioning free energy of proteins as a direct function of membrane curvature, i.e., a curvature sensing profile, within (coarse-grained) molecular dynamics simulations. We demonstrate its utility by resolving these profiles for two known curvature sensing peptides, namely ALPS and α-synuclein, for a membrane curvature ranging from -1/6.5 to +1/6.5 nm-1. We illustrate that the difference in relative partitioning (binding) free energy between these two extrema is only about 13 kBT for both peptides, illustrating that the driving force of curvature sensing is subtle. Furthermore, we illustrate that ALPS and α-synuclein sense curvature via a contrasting mechanism, which is differentially affected by membrane composition. In addition, we demonstrate that the intrinsic spontaneous curvature of both of these peptides lies beyond the range of membrane curvature accessible in micropipette aspiration experiments, being about 1/7 nm -1. Our approach offers an efficient and simple to implement in silico tool for exploring and screening the membrane curvature sensing mechanisms of proteins.

Membrane Thinning Induces Sorting of Lipids and the Amphipathic Lipid Packing Sensor (ALPS) Protein Motif.

Heterogeneities (e.g., membrane proteins and lipid domains) and deformations (e.g., highly curved membrane regions) in biological lipid membranes cause lipid packing defects that may trigger functional sorting of lipids and membrane-associated proteins. To study these phenomena in a controlled and efficient way within molecular simulations, we developed an external field protocol that artificially enhances packing defects in lipid membranes by enforcing local thinning of a flat membrane region. For varying lipid compositions, we observed strong thinning-induced depletion or enrichment, depending on the lipid's intrinsic shape and its effect on a membrane's elastic modulus. In particular, polyunsaturated and lysolipids are strongly attracted to regions high in packing defects, whereas phosphatidylethanolamine (PE) lipids and cholesterol are strongly repelled from it. Our results indicate that externally imposed changes in membrane thickness, area, and curvature are underpinned by shared membrane elastic principles. The observed sorting toward the thinner membrane region is in line with the sorting expected for a positively curved membrane region. Furthermore, we have demonstrated that the amphipathic lipid packing sensor (ALPS) protein motif, a known curvature and packing defect sensor, is effectively attracted to thinner membrane regions. By extracting the force that drives amphipathic molecules toward the thinner region, our thinning protocol can directly quantify and score the lipid packing sensing of different amphipathic molecules. In this way, our protocol paves the way toward high-throughput exploration of potential defect- and curvature-sensing motifs, making it a valuable addition to the molecular simulation toolbox.