Interpretation of Phase Boundary Fluctuation Spectra in Biological Membranes with Nanoscale Organization

Abstract

In this work, we use support vector machine algorithm to detect simple and complex interfaces in atomistic and coarse-grained molecular simulation trajectories of phase-separating lipid bilayer systems. We show that the power spectral density of the interfacial height fluctuations and, in turn, the line tension of the lipid bilayer systems depends on the order parameter used to identify the intrinsic interface. To highlight the effect of artificial smoothing of the interface on the fluctuation spectra and the ensuing line tension calculations, we perform a convolution of the boundaries identified at molecular resolution with a 2D Gaussian function of variance ε2 equal to the resolution limit, (1/2πε2)­exp­( – |r|2/2ε2). The convolution function is given by h ⊗ g where h is the instantaneous height fluctuation, and g is the Gaussian function. This is similar to the effect of point spread functions in experiments. We find that the region of fluctuation spectra that scales according to capillary wave theory formalism depends on the complexity of the interfacial geometry, which may not always be detected at experimental resolutions. We propose that the different k regimes in the fluctuation spectra can be used to characterize mode-dependent interfacial tensions to understand the interfaces beyond the linear line tension calculations.

Enthalpic and entropic contributions to interleaflet coupling drive domain registration and antiregistration in biological membrane

“…Nanoscale molecular-level heterogeneity and structures are nontrivial to quantify [52][53][54][55].We used a depth-first search (DFS) [56] based algorithm to determine domain size distributions for each leaflet for a given configuration. We chose the DFS algorithm because it is computationally very light compared to more involved methods like support vector machine (SVM) [57] analysis for domain boundary and size determination. It also does not require any visual aid in carrying out the categorization, which would be very impractical for such a large number of trajectories.…”

MLLPA: A Machine Learning‐assisted Python module to study phase‐specific events in lipid membranes

“…The interest for ML‐based analysis tool is growing exponentially in all science fields, as it has been widely proven to be a versatile and robust tool that drastically reduce the bias brought by the user in the selection of the parameters to consider 12,13 . ML has been already applied to lipid molecules and membranes to predict nanostructures or predict and locate thermodynamic phases interfaces 14–17 . Other non‐ML‐based approaches to detect phase transition are still being developed, 18 but they remain scarce.…”