Overlay databank unlocks data-driven analyses of biomolecules for all.

Tools based on artificial intelligence (AI) are currently revolutionising many fields, yet their applications are often limited by the lack of suitable training data in programmatically accessible format. Here we propose an effective solution to make data scattered in various locations and formats accessible for data-driven and machine learning applications using the overlay databank format. To demonstrate the practical relevance of such approach, we present the NMRlipids Databank-a community-driven, open-for-all database featuring programmatic access to quality-evaluated atom-resolution molecular dynamics simulations of cellular membranes. Cellular membrane lipid composition is implicated in diseases and controls major biological functions, but membranes are difficult to study experimentally due to their intrinsic disorder and complex phase behaviour. While MD simulations have been useful in understanding membrane systems, they require significant computational resources and often suffer from inaccuracies in model parameters. Here, we demonstrate how programmable interface for flexible implementation of data-driven and machine learning applications, and rapid access to simulation data through a graphical user interface, unlock possibilities beyond current MD simulation and experimental studies to understand cellular membranes. The proposed overlay databank concept can be further applied to other biomolecules, as well as in other fields where similar barriers hinder the AI revolution.

Probing the dynamic landscape of peptides in molecular assemblies by synergized NMR experiments and MD simulations.

Peptides or proteins containing small biomolecular aggregates, such as micelles, bicelles, droplets and nanodiscs, are pivotal in many fields ranging from structural biology to pharmaceutics. Monitoring dynamics of such systems has been limited by the lack of experimental methods that could directly detect their fast (picosecond to nanosecond) timescale dynamics. Spin relaxation times from NMR experiments are sensitive to such motions, but their interpretation for biomolecular aggregates is not straightforward. Here we show that the dynamic landscape of peptide-containing molecular assemblies can be determined by a synergistic combination of solution state NMR experiments and molecular dynamics (MD) simulations. Solution state NMR experiments are straightforward to implement without an excessive amount of sample, while direct combination of spin relaxation data to MD simulations enables interpretation of dynamic landscapes of peptides and other aggregated molecules. To demonstrate this, we interpret NMR data from transmembrane, peripheral, and tail anchored peptides embedded in micelles. Our results indicate that peptides and detergent molecules do not rotate together as a rigid body, but peptides rotate in a viscous medium composed of detergent micelle. Spin relaxation times also provide indirect information on peptide conformational ensembles. This work gives new perspectives on peptide dynamics in complex biomolecular assemblies.

Charged Small Molecule Binding to Membranes in MD Simulations Evaluated against NMR Experiments.

Interactions of charged molecules with biomembranes regulate many of their biological activities, but their binding affinities to lipid bilayers are difficult to measure experimentally and model theoretically. Classical molecular dynamics (MD) simulations have the potential to capture the complex interactions determining how charged biomolecules interact with membranes, but systematic overbinding of sodium and calcium cations in standard MD simulations raises the question of how accurately force fields capture the interactions between lipid membranes and charged biomolecules. Here, we evaluate the binding of positively charged small molecules, etidocaine, and tetraphenylphosphonium to a phosphatidylcholine (POPC) lipid bilayer using the changes in lipid head-group order parameters. We observed that these molecules behave oppositely to calcium and sodium ions when binding to membranes: (i) their binding affinities are not overestimated by standard force field parameters, (ii) implicit inclusion of electronic polarizability increases their binding affinity, and (iii) they penetrate into the hydrophobic membrane core. Our results can be explained by distinct binding mechanisms of charged small molecules with hydrophobic moieties and monoatomic ions. The binding of the former is driven by hydrophobic effects, while the latter has direct electrostatic interactions with lipids. In addition to elucidating how different kinds of charged biomolecules bind to membranes, we deliver tools for further development of MD simulation parameters and methodology.

Headgroup Structure and Cation Binding in Phosphatidylserine Lipid Bilayers.

Phosphatidylserine (PS) is a negatively charged lipid type commonly found in eukaryotic membranes, where it interacts with proteins via nonspecific electrostatic interactions as well as via specific binding. Moreover, in the presence of calcium ions, PS lipids can induce membrane fusion and phase separation. Molecular details of these phenomena remain poorly understood, partly because accurate models to interpret the experimental data have not been available. Here we gather a set of previously published experimental NMR data of C-H bond order parameter magnitudes, |SCH|, for pure PS and mixed PS:PC (phosphatidylcholine) lipid bilayers and augment this data set by measuring the signs of SCH in the PS headgroup using S-DROSS solid-state NMR spectroscopy. The augmented data set is then used to assess the accuracy of the PS headgroup structures in, and the cation binding to, PS-containing membranes in the most commonly used classical molecular dynamics (MD) force fields including CHARMM36, Lipid17, MacRog, Slipids, GROMOS-CKP, Berger, and variants. We show large discrepancies between different force fields and that none of them reproduces the NMR data within experimental accuracy. However, the best MD models can detect the most essential differences between PC and PS headgroup structures. The cation binding affinity is not captured correctly by any of the PS force fields-an observation that is in line with our previous results for PC lipids. Moreover, the simulated response of the PS headgroup to bound ions can differ from experiments even qualitatively. The collected experimental data set and simulation results will pave the way for development of lipid force fields that correctly describe the biologically relevant negatively charged membranes and their interactions with ions. This work is part of the NMRlipids open collaboration project ( nmrlipids.blogspot.fi ).

Accurate Binding of Sodium and Calcium to a POPC Bilayer by Effective Inclusion of Electronic Polarization.

Binding affinities and stoichiometries of Na+ and Ca2+ ions to phospholipid bilayers are of paramount significance in the properties and functionality of cellular membranes. Current estimates of binding affinities and stoichiometries of cations are, however, inconsistent due to limitations in the available experimental and computational methods. In this work, we improve the description of the binding details of Na+ and Ca2+ ions to a 1-palmitoyl-2-oleoyl-phosphatidylcholine (POPC) bilayer by implicitly including electronic polarization as a mean field correction, known as the electronic continuum correction (ECC). This is applied by scaling the partial charges of a selected state-of-the-art POPC lipid model for molecular dynamics simulations. Our improved ECC-POPC model reproduces not only the experimentally measured structural parameters for the ion-free membrane, but also the response of lipid headgroup to a strongly bound cationic amphiphile, as well as the binding affinities of Na+ and Ca2+ ions. With our new model, we observe on the one side negligible binding of Na+ ions to POPC bilayer, while on the other side stronger interactions of Ca2+ primarily with phosphate oxygens, which is in agreement with the previous interpretations of the experimental spectroscopic data. The present model results in Ca2+ ions forming complexes with one to three POPC molecules with almost equal probabilities, suggesting more complex binding stoichiometries than those from simple models used to interpret the NMR data previously. The results of this work pave the way to quantitative molecular simulations with realistic electrostatic interactions of complex biochemical systems at cellular membranes.