Search

Bicontinuous cubic lipidic materialsare increasingly used as crystallizationmedia for in meso crystallization of membrane proteins (MPs). Varyingthe lipid architecture may assist with encapsulation of larger proteinsand promote crystal growth. However, not all lipids are compatiblewith the components of typical crystallization screens, and compatibilitymust therefore be checked prior to crystallization trials. The methodcurrently used, high-throughput small-angle X-ray scattering (HT SAXS),may be time-consuming and is costly in valuable MP. We have thereforeemployed a modeling approach using Bayesian regularized neural networksto accurately predict the complex phase behavior of lipid materialsunder the influence of the PACT crystallization screen and determinethe lipid characteristics that allow a lipid to retain a cubic phaseunder the multiple components required during an in meso crystallizationtrial. This information will be used to select robust lipids for usein crystallization trials and may allow for the rational design ofnew lipids, specifically for in meso crystallization. [ABSTRACT FROM AUTHOR]

Membrane-boundproteins comprise a very important class of drugtargets. Solution of their structures by X-ray crystallography hasbeen hampered by difficulties in crystallizing them in biologicallyrelevant conformations. Novel amphiphilic materials that form bicontinuouscubic phases are being used to support the growth of crystals. Howeverthe cubic phase may transit to other lipidic mesophase structuresunder the influence of the different components within the crystallizationscreen. Furthermore the mesophases may evolve with time, a processthat is poorly understood but potentially critical for controlledcrystal growth. Recent advances in high-throughput screening of lipidsystems have allowed us to generate a large body of data on the influenceof screen components on the cubic phase. However it has been difficultto deconvolute individual effects in the multicomponent system presentduring a crystallization trial. We have therefore developed robustand predictive computational models that predict how the phase behaviorof lyotropic liquid crystals changes over time and under the influenceof crystallization additives. Our work demonstrates that the complexphase behavior of amphiphilic nanostructured nanoparticles can becaptured with high accuracy using modern, robust machine learningmethods. We predicted the existence of individual nanophases withaccuracies of 98–99% and the complex coexistence of multiplephases to a similar accuracy using nonlinear models: linear modelswere not as effective and robust. This approach also allowed us todetermine which crystallization screen components were most relevantto the temporal evolution of individual mesophases. [ABSTRACT FROM AUTHOR]