PREPRINT Machine Learning for the Sensitivity Analysis of a Model of the Cellular Uptake of Nanoparticles for the Treatment of Cancer.

Abstract

Experimental studies on the cellular uptake of nanoparticles (NPs), useful for the investigation of NP-based drug delivery systems, are often difficult to interpret due to the large number of parameters that can contribute to the phenomenon. It is therefore of great interest to identify insignificant parameters to reduce the number of variables used for the design of experiments. In this work, a model of the wrapping of elliptical NPs by the cell membrane is used to compare the influence of the aspect ratio of the NP, the membrane tension, the NP-membrane adhesion, and its variation during the interaction with the NP on the equilibrium state of the wrapping process. Several surrogate models, such as Kriging, Polynomial Chaos Expansion (PCE), and artificial neural networks (ANN) have been built and compared to emulate the computationally expensive model. Only the ANN-based model outperformed the other approaches by providing much better predictivity metrics and could therefore be used to compute the sensitivity indices. Our results showed that the NP's aspect ratio, the initial NP-membrane adhesion, the membrane tension, and the delay for the increase of the NP-membrane adhesion after receptor dynamics are the main contributors to the cellular internalization of the NP, while the influence of other parameters is negligible.