Ecotoxicity prediction by adaptive fuzzy partitioning: comparing descriptors computed on 2D and 3D structures.

Classification models were established on four endpoints, i.e. trout, daphnia, quail and bee, including from 100 to 300 pesticides subdivided into 3 toxicity classes. For each species, five separate sets of molecular descriptors, computed by several software, were compared, including parameters related to 2D or 3D structures. The most relevant descriptors were selected with help of a procedure based on genetic algorithms. Then, structure-activity relationships were built by Adaptive Fuzzy Partition (AFP), a recursive partitioning method derived from Fuzzy Logic concepts.Globally, satisfactory results were obtained for each animal species. The best cross-validation and test set scores reached values of about 70-75%. More important, the relationships derived from the descriptors calculated from 2D structures were superior or similar to those computed from 3D structures. These results underline that the long computational time employed to compute 3D descriptors is often useless to improve the prediction ability of the ecotoxicity models. Finally, the differences in the prediction ability between the different software used were quite reduced and show the possibility to use different descriptor packages for obtaining similar satisfactory models.

Chemometric studies of retention in capillary gas chromatographic separation of hydrocarbons in coupled columns.

This paper describes how different multivariate analysis and classification methods can be used, to characterize the gas chromatographic separation of complex hydrocarbon mixtures in three columns coupled in series. Principal component analysis (PCA), correspondence factor analysis (CFA), and hierarchical ascending classification (HAC) were used as potential tools for evaluating the experiments on single columns and on column series. It has been demonstrated that: (1) multivariate analysis with PCA and CFA offers a powerful strategy to search for the main factors influencing the separation of hydrocarbons without a priori knowledge of the key factors of the separation. (2) With CFA the contribution of retention due to vapour pressure can be minimized. The use of retention indices, which use the n-alkanes as reference compounds, also helps to decrease the dominant focus on vapour pressure in favor of the more selectivity-based interaction forces. (3) CFA helps to analyze the degree of relevance of the chosen experimental design to the most important factors, controlling chromatographic selectivity.

[Application of neural networks using the volume learning algorithm for the study of structure-activity relationship in chemical compounds]. / Primenenie neironnykh setei, ispol'zuiushchikh algoritm prostranstvennogo obucheniia dlia issledovaniia sviazei struktura-aktivnost' khimicheskikh soedinenii.

A volume learning algorithm for artificial neural networks was developed to quantitatively describe the three-dimensional structure-activity relationships using as an example N-benzylpiperidine derivatives. The new algorithm combines two types of neural networks, the Kohonen and the feed-forward artificial neural networks, which are used to analyze the input grid data generated by the comparative molecular field approach. Selection of the most informative parameters using the algorithm helped to reveal the most important spatial properties of the molecules, which affect their biological activities. Cluster regions determined using the new algorithm adequately predicted the activity of molecules from a control data set.

Database mining applied to central nervous system (CNS) activity.

A data set of 389 compounds, active in the central nervous system (CNS) and divided into eight classes according to the receptor type, was extracted from the RBI database and analyzed by Self-Organizing Maps (SOM), also known as Kohonen Artificial Neural Networks. This method gives a 2D representation of the distribution of the compounds in the hyperspace derived from their molecular descriptors. As SOM belongs to the category of unsupervised techniques, it has to be combined with another method in order to generate classification models with predictive ability. The fuzzy clustering (FC) approach seems to be particularly suitable to delineate clusters in a rational way from SOM and to get an automatic objective map interpretation. Maps derived by SOM showed specific regions associated with a unique receptor type and zones in which two or more activity classes are nested. Then, the modeling ability of the proposed SOM/FC Hybrid System tools applied simultaneously to eight activity classes was validated after dividing the 389 compounds into a training set and a test set, including 259 and 130 molecules, respectively. The proper experimental activity class, among the eight possible ones, was predicted simultaneously and correctly for 81% of the test set compounds.

A molecular modeling and 3D QSAR study of a large series of indole inhibitors of human non-pancreatic secretory phospholipase A2.

Automated docking allowing protein-based alignment was performed for a series of 188 indole inhibitors of the human non-pancreatic secretory phospholipase A2 (hnps-PLA2). All the substituted indoles were docked to the crystal structure of hnps-PLA2 and a three-dimensional QSAR model was then established using the CoMFA method. The set of 188 compounds was divided into two subsets, the first one constituting the training set (126 compounds), while the second constituted the test set (62 compounds). The established CoMFA model derived from the training set was then applied to the test set. A good correlation between predicted and experimental activity data allows to validate the 3D QSAR model. A second and global 3D QSAR including all the compounds was established, allowing the creation of the hnps-PLA2 pharmacophore.

Protein-based alignment in 3D QSAR of 26 indole inhibitors of human pancreatic phospholipase A2.

An automated docking procedure was applied on a series of 26 reversible and competitive indole inhibitors of human pancreatic phospholipase A2 (hp-PLA2). X-ray data of this enzyme are not available and the structure was then reconstructed exploiting its protein sequence and the crystallographic data of a bovine pancreatic source. The docking data were used to build a three-dimensional quantitative structure-activity relationship (3D QSAR) model, established using the comparative molecular field analysis (CoMFA) method. This model, joined to the previous one developed for the indole inhibitors of human non-pancreatic secretory phospholipase A2 (hnps-PLA2), an enzyme involved in inflammation processes, will allow for the selection of new strong anti-inflammatory drugs with negligible side effects, at least at the level of hp-PLA2.

Hybrid systems for virtual screening: interest of fuzzy clustering applied to olfaction.

Kohonen neural networks, also known as Self Organizing Map (SOM), offer a useful 2D representation of the compound distribution inside a large chemical database. This distribution results from the compound organization in a molecular diversity hyperspace derived from a large set of molecular descriptors. Fuzzy techniques based on the "concept of partial truth" reveal to be also a valuable tool for the direct exploitation of chemical databases or SOM. In such cases a fuzzy clustering algorithm is used. In this paper, a complete hybrid system, combining SOM and fuzzy clustering, is applied. As example, a series of olfactory compounds was selected. The complexity of such information is that a same compound may exhibit different odors. It is shown how fuzzy logic helps to have a better understanding of the organization of the compounds. These hybrid systems, using simultaneously SOM and fuzzy clustering, are foreseen as powerful tools for "virtual pre-screening".

Validation of protein-based alignment in 3D quantitative structure-activity relationships with CoMFA models.

The predictive capabilities of protein-based alignment (PBA) and structure-based alignment (SBA) comparative molecular field analysis (CoMFA) models have been compared. 3D quantitative structure-activity relationship (3D QSAR) models have been derived for a series of N-benzylpiperidine derivatives which are potent acetylcholinesterase (AChE) inhibitors interesting for Alzheimer's disease. To establish a comparison with the classical SBA procedure, different assay models were derived by superposing ligand conformers that are docked to the AChE active site and by using the most active compound as the reference one. A Kohonen self organizing map (SOM) was applied to analyse the molecular diversity of the test set relative to that of the training set, in order to explain the influence of molecular diversity on the predictive power of the considered models. SBA 3D QSAR models have to be used to predict the inhibitory activity only for compounds belonging to subgroups included in the training set. The PBA 3D QSAR models appeared to have a higher predictability, even for compounds with a molecular diversity greater than that of the training set. This results from the fact that the protein helps to automatically select the active conformation which is fitting the 3D QSAR model.

Automated docking of 82 N-benzylpiperidine derivatives to mouse acetylcholinesterase and comparative molecular field analysis with 'natural' alignment.

Automated docking and three-dimensional Quantitative Structure-Activity Relationship studies (3D QSAR) were performed for a series of 82 reversible, competitive and selective acetylcholinesterase (AChE) inhibitors. The suggested automated docking technique, making use of constraints taken from experimental crystallographic data, allowed to dock all the 82 substituted N-benzylpiperidines to the crystal structure of mouse AChE, because of short computational times. A 3D QSAR model was then established using the CoMFA method. In contrast to conventional CoMFA studies, the compounds were not fitted to a reference molecule but taken in their 'natural' alignment obtained by the docking study. The established and validated CoMFA model was then applied to another series of 29 N-benzylpiperidine derivatives whose AChE inhibitory activity data were measured under different experimental conditions. A good correlation between predicted and experimental activity data shows that the model can be extended to AChE inhibitory activity data measured on another acetylcholinesterase and/or at different incubation times and pH level.

Computer-aided molecular selection and design of natural bioactive molecules.

The search for natural bioactive compounds has led to a renewal of interest in exploring the plant kingdom. Indeed, a more rational search for innovative natural active compounds has become a priority. This review describes the search for new, natural, active compounds by combining the classical ethnopharmacology approach with newer strategies. The proposed computer-aided molecular selection and design (CAMSD) strategy is based on an in-depth exploitation of all the ethnopharmacological, chemical and biological information available. In the first step, the information extracted from various complementary sources - private, literature, Internet - is organized within a database called Phytotech. In the second step, bioinformatic technologies are used to search for new leads based on the molecular and/or the botanical diversity analysis of the Phytotech database. Once a lead is found, the knowledge of involved protein/ligand interaction is improved by molecular modeling. Finally, the activity of the derived bioactive compounds is optimized by pharmacomodulation of the previously selected leads with the help of two- or three-dimensional quantitative structure-activity relationships (2D- or 3D-QSAR) database exploitation. This coherent and global strategy, specially designed for selecting and designing natural bioactive molecules, is based on the hybridization of various chemometric strategies and supported by our own recent examples dealing with acetylcholinesterase inhibition.

Application of Kohonen Neural Networks in classification of biologically active compounds.

Automated data classification is an indispensable tool in Drug Design. It allows to select homogeneous training sets or to distinguish compounds with required biological properties. The Kohonen Neural Networks (KNN) suggest new means for classification of biologically interesting compounds. In this paper, first, capabilities of KNN in data dimensionality reduction are presented as compared with the capabilities of Principal Component Analysis (PCA) and Hierarchical Cluster Analysis (HCA). The advantages of KNN become evident with increasing data dimensionality and size of the training set. Then, new methods are suggested to evaluate the quality of KNN models. Finally, a case study on chemical and biological data is presented. The database studied includes more than 2000 organophosphorous potent pesticides. The Kohonen maps were obtained which allow to distinguish compounds with different biological behavior.

Estimation of blood-brain barrier crossing of drugs using molecular size and shape, and H-bonding descriptors.

The influence of physicochemical properties, including lipophilicity, H-bonding capacity and molecular size and shape descriptors on brain uptake has been investigated using a selection of marketed CNS and CNS-inactive drugs. It is demonstrated that the polar surface area of a drug can be used as a suitable descriptor for the drugs' H-bonding potential. A combination of a H-bonding and a molecular size descriptor, i.e., the major components of lipophilicity and permeability, avoiding knowledge of distribution coefficients, is proposed to estimate brain penetration potential of new drug candidates. Previously reported experimental surface activity data appear to be strongly correlated to molecular size of the drug compounds. Present analysis offers a modern basis for property-based design and targeting of CNS drugs.

A 3D QSAR study of a series of HEPT analogues: the influence of conformational mobility on HIV-1 reverse transcriptase inhibition.

Quantitative structure-activity relationships (QSAR) have been established for 87 analogues of 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT), a potent inhibitor of the HIV-1 reverse transcriptase (RT). Of these 87 nonnucleoside RT inhibitors, 9 novel HEPT analogues were used in the study and the others were taken from the literature. The predictive ability of these relationships has been evaluated using a large set of 54 compounds which were not used to derive the activity model. Descriptors related to the conformational changes were found to be an important factor which underlies RT inhibitory activity in the HEPT series. Indeed, the QSAR model provides evidence concerning the conformational transformations the molecules may undergo during the inhibition process. The established relationships are supplementary to the experimental study on the binding of HEPT type inhibitors to RT by Hopkins et al. (J. Med. Chem. 1996, 39, 1589-1600). The present study suggests a quantitative interpretation of the structure-activity relationships which otherwise cannot be explained within the framework of the crystal inhibitor-protein model. This information is pertinent to the further design of new HEPT type RT inhibitors.