HQSAR allows fragments to be distinguished based on various parameters, which are defined in Table 9.1.

Many HQSAR models have been generated for a variety of ligands of important molecular targets, including cases where the crystal structure of the target protein is not yet available [14–17]. One has to recognize that along with the prediction of potency and affinity of untested compounds, HQSAR models can provide useful insight into the relationships between structural fragments and biological activity. These fragments can be simply visualized through the generation of contribution maps that replicate the individual contribution of each atom or structural fragment for a query compound. This information, combined with synthetic and medicinal chemistry knowledge, could lead to the synthesis of new drug molecules with improved potency. Thus, HQSAR can be considered a flexible tool in drug design.

As an example, we can cite that the most potent compound of a large collection of 9-substituted-9-deazaguanine analogs that inhibit human purine nucleoside phosphorylase (PNP) enzyme has been identified by Castilho et al. [3] with the application of the 2D-HQSAR contribution map. The contribution map allowed the recognition of key structural components essential for intermolecular interactions in the protein active site. Analyzing the contribution map of the most potent inhibitor, the authors concluded that the molecular fragments for the purine ring and cyanil moieties are strongly involved in the inhibitory profile of the compound. Thus, it is possible to identify responsible atomic contributions of a query molecule of increased potency in the absence of the 3D crystal structure of PNP or any other 3D structural information. On the other hand, when 3D structural information is present, this allows a comparison of the HQSAR results to determine if they are in agreement with the 3D chemical environment of the target protein.

HQSAR models have been used in combination with both classical 2D-QSAR and 3D-QSAR methods, suggesting that this method is a valuable tool for lead optimization and drug design process [15–18]. These complementary methods have effectively allowed the integration of additional features in the interpretation of HQSAR models in terms of their chemical and biological significance [19]. As a descriptive example of the amalgamation of QSAR strategies, a study of a large series of flavanoids, dihydrobenzoxathiins, and dihydrobenzodithiins as estrogen receptor (ER) modulators demonstrated that HQSAR and CoMFA can be carried out concurrently to search for a potent drug molecule [4]. The molecular fragments identified by the HQSAR model are strongly correlated to the binding affinity of this series of ligands, which is in accordance with the CoMFA results shown in the steric and electrostatic contour maps. In another study, Avery et al. [20] reported 3D-QSAR (CoMFA) and HQSAR models for a series of 211 artemisinin analogs to develop more potent and less neurotoxic agents for the oral treatment of drug-resistant malaria. The authors found that predictions made with CoMFA and HQSAR models on the test set compounds were in reasonable agreement with the experimentally determined values.

The role of HQSAR in the drug design is evident, and it is enormously vital to realize that the information collected from the patterns of substructural fragments is associated with essential ligand–receptor interactions. We can cite here that in the absence of tubulin-bound discodermolide crystal structures, Salum et al. [21] recently utilized HQSAR to generate molecular recognition patterns that were then combined with molecular modeling studies as a step for the understanding of crucial discodermolide–tubulin interactions associated with the high antiproliferative activity of discodermolide analogs.

HQSAR has been employed to construct models with internal and external robustness, as well as predictivity. The basic approach of HQSAR can be tactically applied in virtual screening (VS) strategies for the identification of hits [22–24]. The analysis of large data sets generated by combinatorial chemistry and high-throughput screening (HTS) techniques has demonstrated the versatility and range of applications of HQSAR. HQSAR includes several desirable characteristics, such as versatility and reliability. The method is suitable for the identification of important structural fragments and for the rapid evaluation and prediction of large chemical libraries, as well as for database mining.

The originality of this method is that it can distinguish the most important structural fragments related to activity and associate them with unique 2D structural characteristics, which can be found in data sets of chemically diverse compounds. Importantly, HQSAR models carry a set of 2D chemical and structural features that can be quantitatively translated into specific 3D ligand–protein interactions. Furthermore, the attribution of diverse weights to differentiate the relative contribution of molecular fragments is directed by experimental values. Statistically robust HQSAR models have been generated for a set of indole derivatives consisting potent anticancer activity for accurate prediction of the activity [25,26].

The computational approach can be employed to search large chemical databases to identify probable hits for a given biological target. Identified molecular recognition patterns from HQSAR models can be combined with the structure-based VS, which is graphically illustrated in Figure 9.4. The integration of fragment-based methods and structure-based VS strategies can be utilized successfully in the coming years.

Along with properties like efficacy, potency, and selectivity, a drug should possess desirable pharmacokinetic/dynamic characteristics and safety. Properties such as absorption, distribution, metabolism, and excretion (ADME) have been recently considered in early phases of drug discovery, as undesirable pharmacokinetic properties can negatively affect the clinical development of new drug molecules [27,28]. In terms of resources, reagents, and detection techniques, in vitro and in vivo ADME assays are quite lengthy, complex, and relatively expensive. In this perspective, a variety of useful in silico ADME models have been developed for the screening of large data sets of compounds, which can be used as faster, simpler, and cheaper tools than the experimental ones [28].

More recently, it has been observed that reliable HQSAR models have been successfully developed and validated for data sets of highly diverse compounds belonging to various therapeutic classes [29,30]. It is interesting to point out that the HQSAR patterns of substructural fragments could also be useful in pharmacokinetic studies comparing traditional mechanism-based pharmacodynamic modeling. The identified substructural patterns for a particular class of chemicals can be employed as ADME filters in the process of chemical library design and VS. The high chemical diversity and the conformation-independent characteristics signify distinctive advantages in the application of the HQSAR method for ADME property prediction.