A new Approach to Active‐Site Binding Analysis. Inhibitors of Acetylcholinesterase

Abstract

Receptor site binding is shown to be localized in selected atoms or substructures with relatively minor contributions from the remaining molecular framework. Related compounds can be overlayed to form a lowest common structure or hyper‐molecule that defines every possible binding site for the series. Any attempt to analyze binding data on an atom‐by‐atom basis would clearly fail if all or most of the positions were involved in the energetics. In most data sets, analysis becomes feasible when most of the variance depends on a minority of the total occupied positions. This appears to be the case in most of the sets analyzed to date, providing us with a unique opportunity. By using atomic descriptors to model lipophilicity, London forces, steric repulsion and charge interactions, the binding site can be mapped by regression analysis. Each important site of interaction and the nature of the binding force can be clearly identified. This procedure provides a statistical method based on measured data that fully complements the visual fitting of modeled drugs into defined receptor sites by computer graphics. It has enormous potential for analysis of drug, agrochemical and toxicological binding data where crystallized enzymes and fully defined sites are unknown and may long remain so. The method is illustrated by examples drawn from acetylcholinesterase receptor site binding.

Traditional topological indexes vs electronic, geometrical, and combined molecular descriptors in QSAR/QSPR research

“…33, No. 6,1993 839 Subset B of electronic, geometrical, and combined descriptors shows only poor correlations for regression models with up to four descriptors included: R2 values belong to the range 0.457-0.685. Even addition of a fifth descriptor provides only fair correlation with R2 of 0.727 (cf.…”

Molecular Field Topology Analysis Method in QSAR Studies of Organic Compounds

“…The MSG structure and the major descriptor contributions (| I j | > 0.05) to activity are shown in Figure . Contrary to ref , we described all the compounds of the series in terms of one model and revealed a certain similarity in the influence of substituents in different positions.…”

“…Inhibition of Acetylcholinesterase by Monosubstituted Phenyl- N -methylcarbamates. One hundred twenty-four compounds of type 1 0 were characterized by their p I 50 values . The optimal model (N F = 6) includes charges (neutral value 0), lipophilicity (neutral value −0.25), and hydrogen-bond acceptor ability (neutral value −2) of atoms as descriptors.…”

“…The positional analysis technique significantly develops these approaches by suggesting the use of the atomic physicochemical parameters that influence ligand binding to the target as descriptors (bond properties are not considered). This method was successfully applied to the analysis of various biological activities, demonstrating the viability of the topological approach taking into account the physicochemical data. , However, in all the examples these parameters were included only for some sites of a hypermolecule even when certain variation of atom types in other sites did exist. Moreover, only the substituent groups were considered, while the common fragment was ignored.…”

Chemoinformatics as a Theoretical Chemistry Discipline

“…[67] In the superstructure-based approach, each individual molecular graph is considered as a subgraph of a common supergraph corresponding to the ensemble of individual graphs. [68] Although this approach is limited to relatively small congeneric sets of compounds, it has been found very suitable to build QSAR models, as demonstrated in the positional analysis by Magee, [69][70] the DARC/CALPHI system by Mercier et al, [71] the MTD-PLS approach of Kurunczi et al, [72][73][74] and the MFTA approach by Palyulin et al [68,75] For each individual chemical structure, the occupancies of supergraph nodes or local physicochemical descriptors of atoms matching these nodes, form a fixed-size descriptor vector used in machine-learning methods as an input.…”