Multiparameter prediction control of in vitro drug delivery into mycobacterium smegmatis induced by microbubble-enhanced sonoporation.

The antibacterial method induced by microbubble-enhanced sonoporation has shown its great potential in facilitating drug delivery into thallus. The enhanced drug delivery induced by microbubble-enhanced sonoporation is a complex event which can be affected by various physical parameters. How to determine the correlation between experimental parameters and the drug delivery efficiency to give the instruction on reasonably choosing the parameters and achieve the control of drug delivery efficiency is impeding further investigations for this complex biophysical process. In the present work, we have explored a number of key parameters affecting the drug delivery efficiency induced by microbubble-enhanced sonoporation using multivariate biological experiments. To achieve the control of the drug delivery efficiency, a multiparameter prediction control method based on modified artificial neural network is presented in this paper. This method is a new modeling method based on combined back-propagation neural network and the multiple model idea to establish quantitative relationship between experimental parameters and drug delivery efficiency. By analyzing the experimental samples, a mapping relationship expression can be deduced to determine the input and output variables of artificial neural network models. Experimental samples were divided into training and test samples. We trained models based on back-propagation neural network to establish their quantitative relationship. In this model, the multiple model idea was introduced into the selection of training samples to modify the traditional back-propagation neural network model to avoid model mismatch caused by poor training sample selection. Numerical experiments results have shown that compared with the traditional model, the identification results obtained by modified model are more closed to experimental results. It is elucidated that an appropriately trained network can act as a good alternative for costly and time-consuming experiments. The findings of this study indicate that this approach can realize the prediction of drug delivery efficiency induced by microbubble-enhanced sonoporation under different experimental parameters, and then achieve the control of drug delivery efficiency through reasonable parameter selection, finally achieve the purpose of efficiently killing bacteria.