Development of classification models for identification of important structural features of isoform-selective histone deacetylase inhibitors (class I).

Abstract

As one of the hot topics in the epigenetic studies, histone deacetylases inhibitors (HDACIs) have been introduced to treat a variety of diseases such as cancer, immune disorder and neuronal diseases. Given the high numbers of available pathways in which HDACs are involved, the HDACIs that act particularly on Class I or Class II enzymes are considered as possible candidates for anticancer drugs. Due to their effective roles in the onset of cancer and its progression, HDAC Class I isoforms (HDAC 1, 2, 3 and 8) were considered in this study. Herein, our objective is to determine the important isoform-selective and isoform-active structural features of HDACIs using the valid classification models. For this purpose, a diverse dataset comprising 8224 HDAC modulators was collected from the binding database. To identify the significant discriminative features, five classification models were generated by supervised Kohonen network and support vector machine methods. Variable importance in projection method was used as a variable selection approach. The results obtained from descriptor analysis show that physicochemical properties, such as hydrogen bonding, number of branches, size, flexibility, polarity and sphericity in the structure of molecules, were closely related to the bioactivity of HDACIs. The reliability and predictive ability of the conducted models were evaluated using the tenfold cross-validation techniques, test sets and applicability domain analysis. All of the obtained classification models represented high statistical quality and predictive ability with accuracy greater than 85% for the test sets. The proposed strategy and the selective patterns represented in this paper can be applied by researchers in the pharmaceutical sciences who aim to use the same idea for the design of drugs with improved anticancer properties.