• Winkler, Dave
• Mulet, Xavier
• Le, Tu

Abstract

When lyotropic liquids are added to a polar solvent, nanocrystals form that adopt a range of morphologies including 1D lamellar structures, 2D inverse hexagonal phases and 3D inverse bicontinuous cubic phases (Fig. 1) The cubic phases in particular have great potential as delivery vehicles for drugs and imaging agents as they are biodegradable, adaptable to multiple drug sizes and types, have enhanced physical and chemical stability, and enhanced tissue and cellular uptake, properties valuable for new therapeutic or diagnostic applications. However little is known about the effect of the incorporated drug structure, properties and loadings on the structure of the resultant cubic nanophases.Figure 1. Structure of the inverse-bicontinuous diamond QIID (Pn3m), gyroid QIIG (Ia3d), and primitive QIIP (Im3m) cubic phases. Computational machine learning tools have been an important molecular design in many areas such as agrochemistry, toxicology and especially pharmaceutical chemistry. We have applied this method to model novel amphiphilic lyotropic liquid crystalline self-assembly materials and to predict their nanomorphology when loaded with drugs. The data was generated using high throughput small angle x-ray scattering (SAXS) at the Australian Synchrotron1. We have developed, for the first time, robust and predictive models that explain the phase behaviour of lipid-based nanoparticles under different conditions. The machine learning methods we used included multiple linear regression and Bayesian regularised artificial neural networks2-4 that generate optimally sparse and predictive models relating structure to properties.In particular, we modelled the phase behaviour of two drug delivery carriers, phytantriol monoolein and Myverol (a commercial product of monoolein) when encapsulating ten drugs at six concentrations at two temperatures. Using state-of-the-art quantitative structure-property relationship modelling techniques, our models successfully predicted the phase formations with very useful accuracy for the first time. We have validated models by predicting the phase behaviour of these drug delivery carriers when loaded with eleven completely new drugs and verifying our predictions using SAXS synchrotron measurements. The model was able to predict a priori the existence of nanophases, or coexistence of nanophases, for new drugs with accuracies of 73-99%Figure 2. Individual phases including inverse-bicontinuous diamond QIID (Pn3m) cubic (blue), hexagonal HII (yellow) and primitive QIIP (Im3m) cubic (purple) predicted by the best BRANN models for Myverol nanoparticles loaded with drugs at 25˚C and 37˚C. Circled samples indicate the mismatch between modelled and experimental results. These robust computational techniques show considerable promise for rationally designing soft nanoparticles for drug delivery, and diverse classes of materials more generally5.1. Mulet, X.; Kennedy, D. F.; Conn, C. E.; Hawley, A.; Drummond, C. J. High throughput preparation and characterisation of amphiphilic nanostructured nanoparticulate drug delivery vehicles. International Journal of Pharmaceutics 2010, 395, 290-297. 2. Burden, F.R.; Winkler, D.A.Robust QSAR Models Using Bayesian Regularized Artificial Neural Networks, J. Med. Chem., 42(16); 3183-3187 (1999).3. Burden, F.R.; Winkler, D.A.An optimal self-pruning neural network that performs nonlinear descriptor selection for QSAR, QSAR Comb. Sci. 28, 1092 – 1097 (2009). 4. Burden, F.R.; Winkler, D.A.Optimum QSAR Feature Selection using Sparse Bayesian Methods, QSAR Comb Sci. 28, 645-653, (2009). 5. Le, T.C.; Epa, V.C.; Burden, F.R.; Winkler, D.A. Towards predictive modelling of diverse materials properties, Chem. Rev. 2012 in press.

Topics

• nanoparticle
• impedance spectroscopy
• phase
• chemical stability
• small angle x-ray scattering
• self-assembly
• machine learning
• lamellae
• gyroid