The Algorithm Of Multiple Linear Regression Model Population Analysis, Implementation And Application In QSAR

In recent years, the method of model population analysis (MPA) has gradually become popular in variable selection and outlier detection. This paper presents a new2D-QSAR modeling method:Multiple Linear Regression Model Population Analysis (MLRMPA), which combines the idea of model population analysis with the multiple linear stepwise regressions. This paper introduced the algorithm of MPA and its application, as well as MLRMPA and its implementation in the paper’s body. Through two application examples, further outline was introduced to discuss parameter setting, analysis the advantages and disadvantages of the algorithm.The first chapter was a review. We primarily focused on the outlines of MPA and MLRMPA. At the same time, R language, as the research tool of MLRMPA algorithm, was also described briefly.The second chapter explored the anti-retroviral activity of RT inhibitors for diarylpyrimidines by MLRMPA method. A comparison was made between the two classes descriptors calculated by Codessa and Dragon softwares to show which one is more suitable for MLRMPA method. A comparison was also made between the two modeling methods (HM and MLRMPA). And we also discussed the parameter setting of MLRMPA in this chapter.The third chapter introduced the MLRMPA package. The main of the chapter contained how to write an R package in the Windows system and how to use the MLRMPA package to analysis chemometrics problems. We combined with an example of how to use MLRMPA package. The modeling results were compared with the random forests’.The last chapter summarized the paper and gave an expectation.