Unbiased Simulations Research Articles

Toward an FPGA-based dedicated computer for molecular dynamics simulations.

First steps toward a molecular dynamics (MD) implementation in a cluster of field-programmable gate arrays (FPGAs) are presented, reaching a simulation speed of a few microseconds/day. The nodes in this cluster are programmed into a mid-ranged FPGA (Artix 7 XC7A200T), interconnected as a 3D torus by fast optical links. The implemented MD algorithm is highly parallelized and highly pipelined internally. The FPGA cluster is freely scalable in terms of size, i.e., a larger MD system requires more nodes, however, without compromising simulation speed. The performance in terms of energy stability and simulation speed is analyzed. At present, the focus lies on the fast networking, while only minimal MD functionality has been implemented so far, i.e., Lennard-Jones interactions and a thermostat, which were needed to demonstrate the feasibility of the FPGA cluster to run multi-microsecond simulations. To that end, the nucleation of a super-cooled Lennard-Jones liquid is investigated by unbiased MD simulations, which is a difficult MD problem since a high nucleation barrier has to be overcome. Finally, the pathways toward a full MD implementation are outlined. The current implementation will be made available as an open-source development project.

Free Energy, Rates, and Mechanism of Transmembrane Dimerization in Lipid Bilayers from Dynamically Unbiased Molecular Dynamics Simulations.

The assembly of proteins in membranes plays a key role in many crucial cellular pathways. Despite their importance, characterizing transmembrane assembly remains challenging for experiments and simulations. Equilibrium molecular dynamics simulations do not cover the time scales required to sample the typical transmembrane assembly. Hence, most studies rely on enhanced sampling schemes that steer the dynamics of transmembrane proteins along a collective variable that should encode all slow degrees of freedom. However, given the complexity of the condensed-phase lipid environment, this is far from trivial, with the consequence that free energy profiles of dimerization can be poorly converged. Here, we introduce an alternative approach, which relies only on simulating short, dynamically unbiased paths, avoiding using collective variables or biasing forces. By merging all paths, we obtain free energy profiles, rates, and mechanisms of transmembrane dimerization with the same set of simulations. We showcase our algorithm by sampling the spontaneous association and dissociation of a transmembrane protein in a lipid bilayer, the popular coarse-grained Martini force field. Our algorithm represents a promising way to investigate assembly processes in biologically relevant membranes, overcoming some of the challenges of conventional methods.

Accelerating Molecular Dynamics through Informed Resetting.

We present a procedure for enhanced sampling of molecular dynamics simulations through informed stochastic resetting. Many phenomena, such as protein folding and crystal nucleation, occur over time scales inaccessible in standard simulations. We recently showed that stochastic resetting can accelerate molecular simulations that exhibit broad transition time distributions. However, standard stochastic resetting does not exploit any information about the reaction progress. For a model system and chignolin in explicit water, we demonstrate that an informed resetting protocol leads to greater accelerations than standard stochastic resetting in molecular dynamics and Metadynamics simulations. This is achieved by resetting only when a certain condition is met, e.g., when the distance from the target along the reaction coordinate is larger than some threshold. We use these accelerated simulations to infer important kinetic observables such as the unbiased mean first-passage time and direct transit time. For the latter, Metadynamics with informed resetting leads to speedups of 2-3 orders of magnitude over unbiased simulations with relative errors of only ∼35-70%. Our work significantly extends the applicability of stochastic resetting for enhanced sampling of molecular simulations.

Solution Thermodynamics of l-Glutamic Acid Polymorphs from Finite-Sized Molecular Dynamics Simulations.

Efficiently obtaining atomic-scale thermodynamic parameters characterizing crystallization from solution is key to developing the modeling strategies needed in the quest for digital design strategies for industrial crystallization processes. Based on the thermodynamics of crystal nucleation in confined solutions, we develop a simulation framework to efficiently estimate the solubility and surface tension of organic crystals in solution from a few unbiased molecular dynamics simulations at a reference temperature. We then show that such a result can be extended with minimal computational overhead to capture the solubility curve. This enables an efficient and self-consistent estimate of the solubility and limit of solution stability associated with crystal nucleation in molecular systems from equilibrium molecular dynamics without the need for sophisticated free energy calculations. We apply our analysis to investigate the relative thermodynamic stability and aqueous solubility of the α and β polymorphs of l-glutamic acid. Our analysis enables an efficient appraisal of emergent ensemble properties associated with the thermodynamics of nucleation from solutions against experimental data, demonstrating that while the absolute solubility is still far from being quantitatively captured by an off-the-shelf point charge transferable force field, the relative polymorphic stability and solubility obtained from finite temperature simulation are consistent with the experimentally available information on glutamic acid. We foresee the ability to efficiently obtain solubility information from a limited number of computational experiments as a critical component of high-throughput polymorph screenings.

Data-driven equation-free dynamics applied to many-protein complexes: The microtubule tip relaxation.

Microtubules (MTs) constitute the largest components of the eukaryotic cytoskeleton and play crucial roles in various cellular processes, including mitosis and intracellular transport. The property allowing MTs to cater to such diverse roles is attributed to dynamic instability, which is coupled to the hydrolysis of guanosine-5'-triphosphate (GTP) to guanosine-5'-diphosphate (GDP) within the β-tubulin monomers. Understanding the dynamics and structural features of both GDP- and GTP-complexed MT tips, especially at an all-atom level, remains challenging for both experimental and computational methods because of their dynamic nature and the prohibitive computational demands of simulating large, many-protein systems. This study employs the "equation-free" multiscale computational method to accelerate the relaxation of all-atom simulations of MT tips toward their putative equilibrium conformation. Using large MT lattice systems (14 protofilaments×8 heterodimers) comprising ∼21-38 million atoms, we applied this multiscale approach to leapfrog through time and nearly double the computational efficiency in realizing relaxed all-atom conformations of GDP- and GTP-complexed MT tips. Commencing from an initial 4 μs unbiased all-atom simulation, we interleave coarse-projective equation-free jumps with short bursts of all-atom molecular dynamics simulation to realize an additional effective simulation time of 1.875 μs. Our 5.875 μs of effective simulation trajectories for each system expose the subtle yet essential differences in the structures of MT tips as a function of whether β-tubulin monomer is complexed with GDP or GTP, as well as the lateral interactions within the MT tip, offering a refined understanding of features underlying MT dynamic instability. The approach presents a robust and generalizable framework for future explorations of large biomolecular systems at atomic resolution.

Thermodynamic origin of fenugreek phytochemical binding to the ASC pyrin domain for inflammation inhibition.

Apoptosis-associated speck-like protein containing a caspase recruitment domain (ASC) is an essential adaptor protein that regulates inflammasome signaling by binding to NOD-like receptor proteins containing pyrin 3 (NLRP3), crucial for the activation of neuroinflammation. Fenugreek phytochemicals are well-known substances with anti-inflammatory properties. The binding interactions of fenugreek phytochemicals toward the ASC pyrin domain (PYD) protein for inflammation inhibition are largely unknown. To this end, all-atom unbiased molecular dynamics simulations that are a total of 58.33 μs long, including absolute binding free energy and umbrella sampling simulations, are performed for unbound and bound ASC-fenugreek complexes. Our calculations reveal that luteolin, one of the flavonoids in fenugreek, previously known to block NLRP3 inflammasome activation and ASC oligomerization, has the strongest binding to ASC among the major phytochemicals, interacting favorably through hydrogen bonds and driven by enthalpic interactions. Electrostatic interactions primarily govern the bindings of flavonoids and alkaloids, whereas van der Waals interactions dictate the binding of saponins. The higher interface water entropy near the binding residues of the ASC-luteolin complex facilitates the binding of luteolin with ASC. The binding residues of ASC toward luteolin, predicted from hydrogen bond occupancy, native contact analyses, and negative binding enthalpy, are similar to those of NLPR3 known from earlier experiments. This suggests that luteolin binding to ASC can potentially block ASC-NLRP3 binding and activation, which regulate inflammatory ASC oligomerization. The study provides insights for governing the binding interactions of fenugreek phytochemicals towards ASC to inhibit inflammasome complex formation, which controls apoptotic signaling pathways, alongside facilitating the drug discovery process for the future.

The importance of sampling the dynamical modes: Reevaluating benchmarks for invariant and equivariant features of machine learning potentials for simulation of free energy landscapes.

Machine learning interatomic potentials (MLIPs) are rapidly gaining interest for molecular modeling, as they provide a balance between quantum-mechanical level descriptions of atomic interactions and reasonable computational efficiency. However, questions remain regarding the stability of simulations using these potentials, as well as the extent to which the learned potential energy function can be extrapolated safely. Past studies have encountered challenges when MLIPs are applied to classical benchmark systems. In this work, we show that some of these challenges are related to the characteristics of the training datasets, particularly the inefficient exploration of the dynamical modes and the inclusion of rigid constraints. We demonstrate that long stability in simulations with MLIPs can be achieved by generating unconstrained datasets using unbiased classical simulations, provided that the important dynamical modes are correctly sampled. In addition, we emphasize that in order to achieve precise energy predictions, it is important to resort to enhanced sampling techniques for dataset generation, and we demonstrate that safe extrapolation of MLIPs depends on judicious choices related to the system's underlying free energy landscape and the symmetry features embedded within the machine learning models.

(Invited) open the Gate: Molecular Dynamics Simulation of Adsorption-Induced Phase Transformations of MOF Nanocrystallites

One of the most intriguing features of Metal-organic frameworks (MOFs) is their ability to respond to external stimuli, like the adsorption of guest molecules, with large changes in the accessible pore volume. Theoretical simulations are a key tool to understand such “gate opening” processes on a molecular level. This is typically done within the approximation of periodic boundary conditions (PBC) and using Grand Canonical methods. Such simulations exclude, however, the interface between the MOF and the gas reservoir and ignore any transport effects, which limits them to the thermodynamic equilibrium. Especially for sensing applications, the kinetics of such processes can become important. Recently, our group started to develop and apply methods to simulate MOF nanocrystallites (NCs) beyond PBC, to circumvent such limitations [1,2,3].In the presentation, recent results on the pore opening of DUT-8(Zn) nanocrystallites in carbon dioxide will be shown. Unbiased molecular dynamics simulations reveal a rather fast phase transformation within a few nanoseconds. The first-order phase transition is initiated by a characteristic pore opening at a corner of the crystallite. The anisotropy of the pillared-layer MOF with its channels along the crystallographic c-axis leads to a specific transport pattern for the guest molecules when the pore opening of the MOF proceeds. The simulations allow an unprecedented view of the actual process, including both the mechanics of the phase transformation of the MOF and the diffusive transport of the guest molecules from the reservoir into the MOF.[1] J. Keupp, R. Schmid, Adv. Theory Simul. 2019, 2, 1900117.[2] L. Schaper, J. Keupp, R. Schmid, Front. Chem. 2021, 9, 757680.[3] L. Schaper, R. Schmid, Commun. Chem. 2023, 6, 233.

Selective Protonation of Catalytic Dyad for γ-Secretase-Mediated Hydrolysis Revealed by Multiscale Simulations.

γ-Secretase plays a crucial role in producing disease-related amyloid-β proteins by cleaving the amyloid precursor protein (APP). The enzyme employs its catalytic dyad containing two aspartates (Asp257 and Asp385) to hydrolyze the substrate by a general acid-base catalytic mechanism, necessitating monoprotonation of the two aspartates for efficient hydrolysis. However, the precise protonation states of the aspartates remain uncertain. In this study, we employed a multiscale computational approach to investigate the dependence of the catalytic efficiency of γ-secretase on the protonation states of its catalytic dyad. Over 200 ms unbiased atomistic simulations of the substrate-enzyme complex reveal diverse orientations of the scissile bond of the bound substrate and accessible structural ensembles of the catalytic dyad with Asp257-Asp385 distances fluctuating between 4 and 10 Å. With a quantum mechanics/molecular mechanics (QM/MM) approach accelerated by enhanced sampling techniques, we find that the first step of the hydrolysis reaction, i.e., the formation of a gem-diol intermediate, experiences a higher reaction barrier by ∼2 kcal/mol when Asp385 is protonated. Furthermore, we find that Arg269 of the enzyme is most likely responsible for this preference of the protonation state: its basic side chain is spatially close to that of Asp257 and specifically stabilizes the transition state electrostatically when Asp257 is protonated. Collectively, our study suggests that Asp257 is likely the favored protonation site for APP cleavage by γ-secretase.

Improved Estimates of Folding Stabilities and Kinetics with Multiensemble Markov Models.

Markov State Models (MSMs) have been widely applied to understand protein folding mechanisms by predicting long time scale dynamics from ensembles of short molecular simulations. Most MSM estimators enforce detailed balance, assuming that trajectory data are sampled at an equilibrium. This is rarely the case for ab initio folding studies, however, and as a result, MSMs can severely underestimate protein folding stabilities from such data. To remedy this problem, we have developed an enhanced-sampling protocol in which (1) unbiased folding simulations are performed and sparse tICA is used to obtain features that best capture the slowest events in folding, (2) umbrella sampling along this reaction coordinate is performed to observe folding and unfolding transitions, and (3) the thermodynamics and kinetics of folding are estimated using multiensemble Markov models (MEMMs). Using this protocol, folding pathways, rates, and stabilities of a designed α-helical hairpin, Z34C, can be predicted in good agreement with experimental measurements. These results indicate that accurate simulation-based estimates of absolute folding stabilities are within reach, with implications for the computational design of folded miniproteins and peptidomimetics.

Allosteric changes in the conformational landscape of Src kinase upon substrate binding

Precise regulation of protein kinase activity is crucial in cell functions, and its loss is implicated in various diseases. The kinase activity is regulated by interconverting active and inactive states in the conformational landscape. However, how protein kinases switch conformations in response to different signals such as the binding at distinct sites remains incompletely understood. Here, we predict the binding mode for the peptide substrate to Src tyrosine kinase using enhanced conformational sampling simulations (totaling 24 μs) and then investigate changes in the conformational landscape upon substrate binding by conducting unbiased molecular dynamics simulations (totaling 50 μs) initiated from the apo and substrate-bound forms. Unexpectedly, the peptide substrate binding significantly facilitates the transitions from active to inactive conformations in which the αC helix is directed outward, the regulatory spine is broken, and the ATP-binding domain is perturbed. We also explore an underlying residue-contact network responsible for the allosteric conformational changes. Our results are in accord with the recent experiments reporting the negative cooperativity between the peptide substrate and ATP binding to tyrosine kinases and will contribute to advancing our understanding of the regulation mechanisms for kinase activity.

Self-organizing maps of unbiased ligand-target binding pathways and kinetics.

The interpretation of ligand-target interactions at atomistic resolution is central to most efforts in computational drug discovery and optimization. However, the highly dynamic nature of protein targets, as well as possible induced fit effects, makes difficult to sample many interactions effectively with docking studies or even with large-scale molecular dynamics (MD) simulations. We propose a novel application of Self-Organizing Maps (SOMs) to address the sampling and dynamic mapping tasks, particularly in cases involving ligand flexibility and induced fit. The SOM approach offers a data-driven strategy to create a map of the interaction process and pathways based on unbiased MD. Furthermore, we show how the preliminary SOM mapping is complementary to kinetic analysis, with the employment of both network-based approaches and Markov state models. We demonstrate the method by comprehensively mapping a large dataset of 640 μs of unbiased trajectories sampling the recognition process between the phosphorylated YEEI peptide and its high-specificity target lck-SH2. The integration of SOM into unbiased simulation protocols significantly advances our understanding of the ligand binding mechanism. This approach serves as a potent tool for mapping intricate ligand-target interactions with unprecedented detail, thereby enhancing the characterization of kinetic properties crucial to drug design.

Statistical Mechanical Theories of Membrane Permeability.

A popular theoretical framework to compute the permeability coefficient of a molecule is provided by the classic Smoluchowski-Kramers treatment of the steady-state diffusive flux across a free-energy barrier. Within this framework, commonly termed "inhomogeneous solubility-diffusion" (ISD), the permeability, P, is expressed in closed form in terms of the potential of mean force and position-dependent diffusivity of the molecule of interest along the membrane normal. In principle, both quantities can be calculated from all-atom MD simulations. Although several methods exist for calculating the position-dependent diffusivity, each of these is at best an estimate. In addition, the ISD model does not account for memory effects along the chosen reaction coordinate. For these reasons, it is important to seek alternative theoretical formulations to determine the permeability coefficient that are able to account for the factors ignored by the ISD approximation. Using Green-Kubo linear response theory, we establish the familiar constitutive relation between the flux density across the membrane and the difference in the concentration of a permeant molecule, j = PΔC. On this basis, we derive a time-correlation function expression for the nonequilibrium flux across a membrane that is reminiscent of the transmission coefficient in the reactive flux formalism treatment of transition rates. An analysis based on the transition path theory framework is exploited to derive alternative expressions for the permeability coefficient. The different strategies are illustrated with stochastic simulations based on the generalized Langevin equation in addition to unbiased molecular dynamics simulations of water permeation of a lipid bilayer.

Exploring the constitutive activation mechanism of the class A orphan GPR20.

GPR20, an orphan G protein-coupled receptor (GPCR), shows significant expression in intestinal tissue and represents a potential therapeutic target to treat gastrointestinal stromal tumors. GPR20 performs high constitutive activity when coupling with Gi. Despite the pharmacological importance of GPCR constitutive activation, determining the mechanism has long remained unclear. In this study, we explored the constitutive activation mechanism of GPR20 through large-scale unbiased molecular dynamics simulations. Our results unveil the allosteric nature of constitutively activated GPCR signal transduction involving extracellular and intracellular domains. Moreover, the constitutively active state of the GPR20 requires both the N-terminal cap and Gi protein. The N-terminal cap of GPR20 functions like an agonist and mediates long-range activated conformational shift. Together with the previous study, this study enhances our knowledge of the self-activation mechanism of the orphan receptor, facilitates the drug discovery efforts that target GPR20.

Molecular Mechanism of Ciprofloxacin Translocation Through the Major Diffusion Channels of the ESKAPE Pathogens Klebsiella pneumoniae and Enterobacter cloacae.

Experimental studies on the translocation and accumulation of antibiotics in Gram-negative bacteria have revealed details of the properties that allow efficient permeation through bacterial outer membrane porins. Among the major outer membrane diffusion channels, OmpF has been extensively studied to understand the antibiotic translocation process. In a few cases, this knowledge has also helped to improve the efficacy of existing antibacterial molecules. However, the extension of these strategies to enhance the efficacy of other existing and novel drugs require comprehensive molecular insight into the permeation process and an understanding of how antibiotic and channel properties influence the effective permeation rates. Previous studies have investigated how differences in antibiotic charge distribution can influence the observed permeation pathways through the OmpF channel, and have shown that the dynamics of the L3 loop can play a dominant role in the permeation process. Here, we perform all-atom simulations of the OmpF orthologs, OmpE35 from Enterobacter cloacae and OmpK35 from Klebsiella pneumoniae. Unbiased simulations of the porins and biased simulations of the ciprofloxacin permeation processes through these channels provide insight into the differences in the permeation pathway and energetics. In addition, we show that similar to the OmpF channel, antibiotic-induced dynamics of the L3 loop are also operative in the orthologs. However, the sequence and structural differences, influence the extent of the L3 loop fluctuations with OmpK35 showing greater stability in unbiased runs and subdued fluctuations in simulations with ciprofloxacin.

BAD-NEUS: Rapidly converging trajectory stratification.

An issue for molecular dynamics simulations is that events of interest often involve timescales that are much longer than the simulation time step, which is set by the fastest timescales of the model. Because of this timescale separation, direct simulation of many events is prohibitively computationally costly. This issue can be overcome by aggregating information from many relatively short simulations that sample segments of trajectories involving events of interest. This is the strategy of Markov state models (MSMs) and related approaches, but such methods suffer from approximation error because the variables defining the states generally do not capture the dynamics fully. By contrast, once converged, the weighted ensemble (WE) method aggregates information from trajectory segments so as to yield unbiased estimates of both thermodynamic and kinetic statistics. Unfortunately, errors decay no faster than unbiased simulation in WE as originally formulated and commonly deployed. Here, we introduce a theoretical framework for describing WE that shows that the introduction of an approximate stationary distribution on top of the stratification, as in nonequilibrium umbrella sampling (NEUS), accelerates convergence. Building on ideas from MSMs and related methods, we generalize the NEUS approach in such a way that the approximation error can be reduced systematically. We show that the improved algorithm can decrease the simulation time required to achieve the desired precision by orders of magnitude.

AlphaFold-SFA: Accelerated sampling of cryptic pocket opening, protein-ligand binding and allostery by AlphaFold, slow feature analysis and metadynamics.

Sampling rare events in proteins is crucial for comprehending complex phenomena like cryptic pocket opening, where transient structural changes expose new binding sites. Understanding these rare events also sheds light on protein-ligand binding and allosteric communications, where distant site interactions influence protein function. Traditional unbiased molecular dynamics simulations often fail to sample such rare events, as the free energy barrier between metastable states is large relative to the thermal energy. This renders these events inaccessible on the timescales typically simulated by unbiased molecular dynamics, limiting our understanding of these critical processes. In this paper, we proposed a novel unsupervised learning approach termed as slow feature analysis (SFA) which aims to extract slowly varying features from high-dimensional temporal data. SFA trained on small unbiased molecular dynamics simulations launched from AlphaFold generated conformational ensembles manages to capture rare events governing cryptic pocket opening, protein-ligand binding, and allosteric communications in a kinase. Metadynamics simulations using SFA as collective variables manage to sample 'deep' cryptic pocket opening within a few hundreds of nanoseconds which was beyond the reach of microsecond long unbiased molecular dynamics simulations. SFA augmented metadynamics also managed to capture conformational plasticity of protein upon ligand binding/unbinding and provided novel insights into allosteric communication in receptor-interacting protein kinase 2 (RIPK2) which dictates protein-protein interaction. Taken together, our results show how SFA acts as a dimensionality reduction tool which bridges the gap between AlphaFold, molecular dynamics simulation and metadynamics in context of capturing rare events in biomolecules, extending the scope of structure-based drug discovery in the era of AlphaFold.

Structural Insights into Endostatin-Heparan Sulfate Interactions Using Modeling Approaches.

Glycosaminoglycans (GAGs) play a key role in a variety of biological processes in the extracellular matrix (ECM) via interactions with their protein targets. Due to their high flexibility, periodicity and electrostatics-driven interactions, GAG-containing complexes are very challenging to characterize both experimentally and in silico. In this study, we, for the first time, systematically analyzed the interactions of endostatin, a proteolytic fragment of collagen XVIII known to be anti-angiogenic and anti-tumoral, with heparin (HP) and representative heparan sulfate (HS) oligosaccharides of various lengths, sequences and sulfation patterns. We first used conventional molecular docking and a docking approach based on a repulsive scaling-replica exchange molecular dynamics technique, as well as unbiased molecular dynamic simulations, to obtain dynamically stable GAG binding poses. Then, the corresponding free energies of binding were calculated and the amino acid residues that contribute the most to GAG binding were identified. We also investigated the potential influence of Zn2+ on endostatin-HP complexes using computational approaches. These data provide new atomistic details of the molecular mechanism of HP's binding to endostatin, which will contribute to a better understanding of its interplay with proteoglycans at the cell surface and in the extracellular matrix.

Self-Assembly of Model Three- and Four-Patch Colloidal Particles in Two Dimensions.

A coarse-grained effective solvent model of two-patch particles is extended to study the self-assembly of three- and four-patch particles to two-dimensional honeycomb and square lattices, respectively. Employing this model, grand canonical ensemble simulations are done to calculate vapor-liquid equilibria and the critical temperatures for patchy particles of various patch widths. The range of stability of the liquid, although very limited compared to isotropic particles, which interact through a longer-range potential, depends on the patch width and on the number of patches. Biased sampling and unbiased simulations are also done to investigate the mechanism of nucleation and crystal growth for honeycomb and square lattices, self-assembled from three- and four-patch particles, respectively. A two-step mechanism governs the nucleation of both lattices. In the first step, the particles form a dense amorphous network, and in the second step, the particles inside the amorphous network reorient to form crystalline nuclei. Barrier heights for the nucleation of honeycomb and square lattices are 7.8 kBT and 7.4 kBT, which are close to the reported values for the nucleation of the kagome lattice. In agreement with confocal microscopy experiments, the self-assembly in a honeycomb lattice involves the formation of 5- to 7-membered rings. The 5- and 7-membered rings hamper the nucleation of the honeycomb lattice, through defect formation and rotation of the symmetry planes of crystals that form at their surfaces. With the progress of self-assembly, a substantial amount of restructuring of the defects and crystals in their vicinity is needed to heal the defects.

Unbiased MD simulations identify lipid binding sites in lipid transfer proteins.

The characterization of lipid binding to lipid transfer proteins (LTPs) is fundamental to understand their molecular mechanism. However, several structures of LTPs, and notably those proposed to act as bridges between membranes, do not provide the precise location of their endogenous lipid ligands. To address this limitation, computational approaches are a powerful alternative methodology, but they are often limited by the high flexibility of lipid substrates. Here, we develop a protocol based on unbiased coarse-grain molecular dynamics simulations in which lipids placed away from the protein can spontaneously bind to LTPs. This approach accurately determines binding pockets in LTPs and provides a working hypothesis for the lipid entry pathway. We apply this approach to characterize lipid binding to bridge LTPs of the Vps13-Atg2 family, for which the lipid localization inside the protein is currently unknown. Overall, our work paves the way to determine binding pockets and entry pathways for several LTPs in an inexpensive, fast, and accurate manner.