In-vivo Drug Metabolism Prediction Based On Kernel Principal Component And Support Vector Machine

Prediction of drug clearance in vivo, can not only reduce the risk and cost of clinical trials, but also achieve rapid batch screening of new drugs, shorten the research period of new drug development and reduce development costs which has great theoretical and practical significance. Kernel principal component analysis is a new technology of nonlinear principal component extraction, which achieves nonlinear change by a kernel function and is good at feature extraction and noise processing. Support vector machine is developed based on statistical learning theory of VC dimension and structural risk minimization. It has great effect on small sample size problem of pattern recognition and function approximation. This paper is organized as follows:In section 1, status of drug clearance in vivo has been described in detail after reading lots of papers at home and abroad, and kernel principal component and support vector machine algorithm are also described simply. In section 2, clearance of the drug model in vitro and in vivo was studied in detail. The experiment data from the model of liver microsomes has been taken as the source of model input parameters after considering the characteristics of in vivo models, kernel principal component and support vector machine algorithm. In section 3, a weighting function of kernel principal component analysis algorithm, which can overcome the extreme point sensitive problem of small samples, was proposed after in-depth study of kernel principal component analysis. Experiments show that the improved algorithm has a good point on the interference robustness. In addition, error function theoretical analysis shows that the algorithm can be a good solution to the convergence problem. Besides, the improved algorithm was also used in liver microsomes in vitro experiment. The result shows that distance-based kernel principal component analysis algorithm can separate the samples by pH effectively. In section 4, first analysis and deduction of support vector machine algorithm was presented in detail. Then clearance of the drug was predicted based on the two-dimensional component analysis data gotten from kernel principal component analysis algorithm. The result shows that the predicted result based on improved kernel principal component analysis algorithm is not only much better than the result based on traditional method, but also better than the result directly based on support vector machine algorithm.This paper presents an algorithm based on kernel principal component analysis and support vector machine. The predicted result based on this algorithm is satisfactory. As the increasing of training sample this algorithm will have a better application prospect.