Résumé: A simple linear model was used to investigate the correlation between the n-octanol/water partition coefficient (kow) of non-substituted fused polycyclic aromatic hydrocarbons (PAHs). Among (74) 3D-geometrically tested descriptors calculated using the Dragon software, volume V turned out to be the best descriptor to model the considered endpoint (with a squared correlation coefficient (R2) of 0.9844 and a standard error of estimation (s) of 0.132 log units). The correlation coefficient cross-validation (Q2) between experimental and predicted log kow for training and test sets was 0.9811 (for training set) and 0.9828 (for test set), respectively. The reliability of the proposed model was further illustrated using various evaluation techniques: leave-5-out cross-validation, bootstrap, randomization tests, and validation through the test set.

Résumé: The search for an eco-freindly larvicide suitable for vector control requires a budget and considerable time to carry out experiments. Fortunately, the use of QSAR modeling allows the prediction of larvicidal activity of structurally diverse chemicals against mosquitoes in a way quick and costless. This approach can be helpful to study for making biolarvicide with highest ability to destroy mosquito larvae. We propose a QSAR model using two different statistical methods, multiple linear regression (MLR) and Support vector machine (SVM) for predicting the larvicidal activity of 30 compounds of essential oil (EOs) isolated from the root of Asarum heterotropoides against Culex pipiens pallens. A model with four theoretical descriptors derived from Dragon software was developed applying the genetic algorithm (GA)-variable subset selection (VSS) procedure. The statistical parameters, R 2 = 0.9716, Q 2 LOO = 0.9595, s = 0.1690 of the model developed by MLR showed a good predictive capability for log LC50 values. The comparison between the results of SVM and MLR models showed that the SVM model present a good alternative to construct a QSAR model for the prediction of the larvicidal activity.

Résumé: The medicinal plants play a great role in the treatment of a lot of different diseases. The use of remedies and the therapeutic qualities of plants are gaining renewed interest thanks to the improvement of extraction techniques and advances in structural analysis methods for the discovery of new active substances. The objective of our study was designed for the determination of micro constituents polyphenols (pericarps 10.13±0.46 GAE mg/g DM, almonds 9.12±0.26 GAE mg/g DM), and macronutrients sugars (almonds 1.84±0.08 mg/g DM), pericarps (0.494±0.02 mg/g DM), proteins (almonds 3.82±0.04 g/100g DM pericarps 1.74± 0.01 g/100g DM), present in oak cork fruits, and which play a very important role. We were interested in determining their content in dried, powdered oak cork fruits (acorns). The carried out phytochemical screening revealed the presence of flavonoids, tannins, and saponins, the presence of these compounds attribute to this species several therapeutic and pharmacological characteristics. The extraction of secondary and primary metabolites by the use of organic solvents gave significant portions that testify the richness of this species in active ingredients. This species (oak cork fruits) should be put in valor as well as complement or food and making the parapharmaceutical products.

Résumé: Retention indices of 38 benzene derivatives, separated by gas chromatography were correlated with 2 connectivity indices, using PM3 semi-empirical calculation method and hybrid genetic algorithms/ multiple linear regression approach. For the sake of external validation, the available set of chemicals was separated using Kennard and Stone algorithm into training set of 28 compounds and an external set of 10 compounds. The proposed hybrid model was validated using different criterions, and its predictive capability meets the conditions defined by Golbraikh et al. In comparison to the previously published model, our model exhibits a large enhancement and its mechanistic interpretation was attempted to connect the selected variables to the retention phenomenon.

Résumé: A quantitative structure property relationship (QSPR) study is developed using Genetic Algorithm (GA) / Multiple Linear Regression (MLR) for modeling the flash points of 173 unsaturated hydrocarbons, using theoretical molecular descriptors derived from DRAGON software. The studied dataset was randomly separated into two independent subsets: a training set of 139 compounds to build the model and a test set of the removed 34 compounds to validate its predictive ability. The selection of a minimum set of meaningful descriptors was carried out using Genetic Algorithm in the MOBYDIGS Todeschini software. An MLR model of 4 descriptors with a high predictive power was developed for the prediction of the flash points of unsaturated hydrocarbons.The predictive ability of the obtained model was verified using a set of criteria according to Golbraikh and tropsha and its applicability domain was studied using Willians plot. Keywords: Flash point; Unsaturated hydrocarbons; Multiple linear regression; Quantitative structure-property relationship; Model prediction

Résumé: Les facteurs acentriques de 18 composés hydroxylés (alcools, phénols), ont été corrélés linéairement avec 2 descripteurs moléculaires de type géométrique sélectionnés par algorithme génétique, parmi plus de 1600 calculés en utilisant le logiciel de modélisation moléculaire DRAGON. Les différentes statistiques établies (coefficient de détermination multiple et de prédiction; racines des erreurs quadratiques moyennes; test de randomisation) montrent la qualité, la robustesse et les bonnes capacités prédictives internes du modèle construit. Aucune observation aberrante ou influente n’a été relevée.

Résumé: A quantitative structure- property relationship (QSPR) was performed for the prediction of the aqueous solubility of pesticides belonging to four chemical classes: acid, urea, triazine, and carbamate. The entire set of 77 pesticides was divided into a training set of 58 pesticides and a test set of 19 pesticides according to the Snee technique. A six descriptor model, with squared correlation coefficient (R2) of 0.8895 and standard error of estimation (s) of 0.52 log unit, was developed by applying multiple linear regression analysis using the ordinary least square regression method and genetic algorithm- variable subset selection. The reliability of the proposed model was further illustrated using various evaluation techniques: leave- one- out cross- validation, bootstrap, randomization tests, and validation through the test set Keywords pesticides- aqueous solubility- QSPR- molecular descriptors- multiple linear regression

Résumé: quantitative structure–retention relationship (QSRR) study was performed to correlate descriptors representing molecular structures to the Kováts retention indices of polycyclic aromatic hydrocarbon compounds. The complete set of 37 compounds was divided randomly into a training set of 26 compounds and a test set of 11 compounds. In the pool of descriptors calculated using Dragon and Hyperchem software, the modified Harary index

Résumé: The relative toxicity of 48 anilines using the Tetrahymena pyriformis population growth characteristics (concentration causing 50% growth inhibition), available in the literature, was studied. At first, the entire data set was randomly split into a training set (31 chemicals) used to establish the QSAR model, and a test set (17 chemicals) for statistical external validation. A biparametric model was developed using, as independent variables, 3D theoretical descriptors derived from DRAGON software.