The applications of PCA in QSAR studies: A case study on CCR5 antagonists

Abstract

Principal component analysis (PCA), as a well-known multivariate data analysis and data reduction technique, is an important and useful algebraic tool in drug design and discovery. PCA, in a typical quantitative structure-activity relationship (QSAR) study, analyzes an original data matrix in which molecules are described by several intercorrelated quantitative dependent variables (molecular descriptors). Although extensively applied, there is disparity in the literature with respect to the applications of PCA in the QSAR studies. This study investigates the different applications of PCA in QSAR studies using a dataset including CCR5 inhibitors. The different types of preprocessing are used to compare the PCA performances. The use of PC plots in the exploratory investigation of matrix of descriptors is described. This work is also proved PCA analysis to be a powerful technique for exploring complex datasets in QSAR studies for identification of outliers. This study shows that PCA is able to easily apply to the pool of calculated structural descriptors and also the extracted information can be used to help decide upon an appropriate harder model for further analysis.

Low dimensional representations along intrinsic reaction coordinates and molecular dynamics trajectories using interatomic distance matrices

“…[15][16][17] In analytical chemistry, PCA is central in the development of quantitative structure activity relationship (QSAR) models, of particular utility in the pharmaceutical industry. [18][19][20][21] Perhaps most closely related to this study is the use of PCA in computational biology, to capture essential motions of a protein in MD simulations. [22][23][24] There are, however, still some key limitations of PCA: rst, it is assumed that the relationships between features of the data points are linear.…”

Finding Needles in a Haystack: Determining Key Molecular Descriptors Associated with the Blood‐brain Barrier Entry of Chemical Compounds Using Machine Learning

“…When the combined set of 777 descriptors the AUC was 0.82 which indicates that the increase in the number of descriptors did not make any significant improvement in model quality. The strong mutual intercorrelation of many descriptors that is common in QSAR problems may explain such results.…”

“…While there are a host of machine learning methods available and widely used for predictive modelling, the measures of variable importance used for them are different and often empirical . Because of the high degree of collinearity exhibited by QSAR descriptors, it is likely that multiple methods would lead to very different sets of important features being selected for each individual method. This makes the discussion on outputs and possible mechanistic interpretations (as given in Section 3.1) difficult.…”

HIV‐1 Entry Inhibitors: A Review of Experimental and Computational Studies

“…Different uses of principle component analysis (PCA), as a helpful mean for data reduction in multivariate data analysis, have also been the subject of study using a dataset of CCR5 inhibitors by the same research group in 2018. The authors have listed various benefits and proved the usefulness of PCA in QSAR studies …”

“…The authors have listed various benefits and proved the usefulness of PCA in QSAR studies. [61] 3.3. CXCR4 Inhibitors CXCR4 antagonists are divided into three groups: bicyclams, monocyclam, and non-cyclams.…”