The Influence of Lipophilicity on the Classification of Antitumor Acridinones Evaluated by Principal Component Analysis

Marcin      Koba; Tomasz      Baczek

doi:10.2174/1573412911208020157

Abstract

Principal component analysis (PCA) has been performed for some acridinone derivatives with antitumor activity to model relationships between structural descriptors and lipophilicity and their antileukemia activity. Principal component analysis obtained with the use of parameters of lipophilicity led to extract two main factors with eigenvalue higher than 1. The first principal component (factor 1) accounted for, by 84.3% of the variance in data and the second principal component (factor 2) explained 10.0% of data variance indicating that total data variance at the level 94.5% can be explained by the first two principal components. On the other hand, principal component analysis obtained with the use of parameters of lipophilicity together with some other molecular descriptors led to extract five main factors with eigenvalue higher than 1. In this case, the first principal component (factor 1) accounted for, by 62.64% of the variance in data, the second principal component (factor 2) explained 19.04% and the third principal component (factor 3) explained 8.49% of data variance indicating that total data variance at the level 90.17% can be explained by the first three principal components. Moreover, one of the most significant influences on the value of factors possessed parameters describing lipophilicity. However, some other descriptors considered in the study (some of constitutional, topological, molecular and geometrical one) could play an important role in the reflection of acridinone properties, too. More importantly, distribution of individual drugs on the plane determined by two and three principal components produced patterns in good agreement with their chemical structures as well as with their antitumor activity.

Keywords: Acridinones, Antitumor activity, Compounds classification, Data variance, Lipophilicity, Molecular modeling, Principal component analysis (PCA), Statistical analysis, Structural descriptors, Varimax rotation