“In Silico” Design of Potential Anti-HIV Actives Using Fragment Descriptors

Abstract

Substructural Molecular Fragments (SMF) method was applied for computer-aided design of new compounds potentially possessing high anti-HIV activities: tetrahydroimidazobenzodiazepinone (TIBO) derivatives and 1-[2-hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT) derivatives. Using available experimental data, the SMF method was first applied to build QSAR models based on fragment descriptors (atom/bond sequences and "augmented atoms"). The focused virtual combinatorial libraries containing 891 TIBO derivatives and 2640 HEPT derivatives were then generated systematically attaching selected substituents to corresponding Markush structures. Finally, a screening of those libraries using developed QSAR models led to several hits potentially possessing high anti-HIV activity.

New Approach for Accurate QSPR Modeling of Metal Complexation: Application to Stability Constants of Complexes of Lanthanide Ions Ln3+, Ag+, Zn2+, Cd2+ and Hg2+ with Organic Ligands in Water

“…[23,30,33,46,73] MLR is applied to build linear relationships between independent variables (SMF descriptors: X i , i =1, 2,…) and a dependent variable (here target property Y = logK): Y = c 0 + Σc i X i , where every descriptor value (SMF count x ij , j = 1, 2,…, n; here n is the number of ligands) is associated with observed property value (y j , j = 1, 2,…, n), c i is descriptor contribution, and c 0 is the independent term which is omitted in a part of models. The Singular Value Decomposition method is used to fit contributions c i and to minimize the sum of squared residuals which are squared differences between the property values calculated by the model (y j , calc ) and observed values (y j , exp ) in the training set.…”

“…One consensus model combines predictions issued from a multitude of individual models originated from different types of the SMF descriptors and variable selection algorithms. [23,28,35,73,74] Thus for every compound from the test set, the target property is computed as an arithmetic mean of values obtained by individual models excluding those leading to outlying values according to Tompson's rule. [75] If a test compound is identified as being outside an applicability domain (AD) of individual model, the prediction by given model for a given compound is not included in CM.…”

Stability constants of complexes of Zn2+, Cd2+, and Hg2+ with organic ligands: QSPR consensus modeling and design of new metal binders

“…2D structures of the ligands, names of the metal ions as well as corresponding experimental log K values were converted by the EdiSDF data manager [24,26,29] into Structure Data Files (SDF) readable by the MLR and PRM programs of the ISIDA package [30,31]. If several values of the stability constant log K were available for a particular ligand, for selections we followed the recommendations of IUPAC [32]; in some cases the most recent data or the data consistent with respect to different experimental methods were chosen.…”

“…CM combines the predictions issued from many individual models originated from different types of the SMF descriptors. CM allows one to smooth inaccuracies of individual models and ensures more reliable predictions [14,24,31,33]. Thus for each compound from the test set, the program computes the property as an arithmetic mean of values obtained with a collection of selected on training stage individual models excluding those leading to outlying values according to Tompson's rule and a method of ranked series [34], and taking into account an applicability domain (AD) of each model.…”

“…The ISIDA (In SIlico design and Data Analysis) program package [29][30][31] (http://infochim.u-strasbg.fr/ recherche/Download/Download.php) has been used for structure-property modeling, which was performed using two machine learning methods: Multiple Linear Regression Analysis (MLR) [12] and Partial Robust M-regression Algorithm (PRM).…”