Influence of feature rankers in the construction of molecular activity prediction models

In the construction of activity prediction models, the use of feature ranking methods is a useful mechanism for extracting information for ranking features in terms of their significance to develop predictive models. This paper studies the influence of feature rankers in the construction of molecular activity prediction models; for this purpose, a comparative study of fourteen rankings methods for feature selection was conducted. The activity prediction models were constructed using four well-known classifiers and a wide collection of datasets. The ranking algorithms were compared considering the performance of these classifiers using different metrics and the consistency of the ranked features.     

Performance evaluation of the GastroPlusTM software tool for prediction of the toxicokinetic parameters of chemicals

ABSTRACT

The accurate prediction of toxicokinetic parameters arising from oral, dermal and inhalation routes of chemical exposure is a key element in chemical safety assessments. In this research, the physiologically based pharmacokinetic (PBPK) GastroPlus^TM software was evaluated against a series of chemicals for the prediction of toxicokinetic parameters. Overall, 67% of predicted intrinsic clearance (Clint) values were within 1- to 10-fold of empirical data for 463 compounds, and 87% of the predicted fraction unbounded in plasma (Fup) values were 1- to 3-fold of empirical data for 441 compounds. The r² (coefficient of determination) of predicted C_max (maximum plasma concentration) and AUC (Area Under Curve) values versus the corresponding empirical values from oral, inhalation and dermal exposures ranged from 0.04 to 0.92. Among the three exposures, the highest r² values, ranging from 0.80 to 0.92, were observed for oral exposure predictions, where 88% of the compounds had 1- to 10-fold differences between predicted and empirical values for C_max and AUC. The predicted plasma C_ss (steady-state plasma concentration) values were consistent with those C_ss values calculated by in vitro-to-in vivo extrapolation (IVIVE) approaches using experimental parameters. Based on the evaluation results, GastroPlus? can be used as a QSAR/PBPK tool for toxicokinetic parameter predictions.     

Cross-target view to feature selection: identification of molecular interaction features in ligand-target space

There is growing interest in computational chemogenomics, which aims to identify all possible ligands of all target families using in silico prediction models. In particular, kernel methods provide a means of integrating compounds and proteins in a principled manner and enable the exploration of ligand-target binding on a genomic scale. To better understand the link between ligands and targets, it is of fundamental interest to identify molecular interaction features that contribute to prediction of ligand-target binding. To this end, we describe a feature selection approach based on kernel dimensionality reduction (KDR) that works in a ligand-target space defined by kernels. We further propose an efficient algorithm to overcome a computational bottleneck and thereby provide a useful general approach to feature selection for chemogenomics. Our experiment on cytochrome P450 (CYP) enzymes has shown that the algorithm is capable of identifying predictive features, as well as prioritizing features that are indicative of ligand preference for a given target family. We further illustrate its applicability on the mutation data of HIV protease by identifying influential mutated positions within protease variants. These results suggest that our approach has the potential to uncover the molecular basis for ligand selectivity and off-target effects.     

Synthesis of nano-sized zeolite-Y functionalized with 5-amino-3-thiomethyl 1H-pyrazole-4-carbonitrile for effective Fe(III)-chelating strategy

Zeolite crystals having faujasite-type (FAU) topology in the nanometer range were first synthesized from amorphous rice husk ash at a low temperature of 363 K under autogenous pressure. Following this, surface functionalization of the produced zeolite with 5-amino-3-thiomethyl 1H-pyrazole-4-carbonitrile (pyrazole; Py) was carried out by two different methods, namely liquefied-period adsorption of Py (Py/Y_im) and a flexible ligand method (Py/Y_ss). The latter provides a larger amount of Py formed into as-made zeolite-Y. The sorption of Fe(III) onto Py/NaY afforded large meso–macroporosity introduced by the aggregation–assembly between Fe(III)Py complexes and NaY zeolite, which was typically evidenced for Fe(III)Py/Y_ss. The materials were characterized by XRD, FT-IR spectroscopy, thermal analysis (TGA) and porous structure by N₂ adsorption–desorption at 77 K. The XRD evaluation showed that the zeolite structure was managed right after adding Fe(III) to Py/Y, although a change in intensity of the zeolite reflections on complex formation was noticed. The FT-IR spectrum of Fe(III)Py/Y_ss exhibited two bands at 3594 and 3542 cm^?1 assigned to bridging hydroxyl groups associated with a Brönsted site, which did not exist in the spectra of Fe(III)Py/Y_im and Fe(III)-exchanged as-made NaY zeolite. This effect was ascribed to the induced greater electronegativity of the ligand towards Fe(III) species in dissociation of water molecules, producing acidic protons that are potential Brönsted acid sites. It was also evident that the Fe(III) adsorption capacity on Py/Y_ss is greater than on as-made NaY zeolite and Py/Y_im, owing most likely to the increasing concentration of the incorporating Py ligand which leads to an increase in the number of binding sites. The Fe(III) adsorption onto Py/Y_ss was well described by the pseudo-second-order kinetic model. Density functional theory (B3LYP/6-311G*) was performed to understanding the interaction mode of the ligand and complex with zeolite. The QSPR was calculated depending on the optimization geometries, frontier molecular orbitals, thermodynamic parameters, and global chemical reactivates, which were discussed for the studied compounds. The HOMOs, LUMOs and molecular electrostatic potentials were plotted to elucidate the interaction manner of the tested compounds with the zeolite. The nonlinear optical properties were elucidated via 1st and 2nd hyper-polarizabilities. The auto-degradation behavior was predicted for the complex, based on the ionization optional and bond dissociation enthalpy. The interactions between P_y and Fe(III)P_y with the zeolite surface have been described with molecular dynamics using a Monte Carlo simulation.     

Interpretation of Ligand-Based Activity Cliff Prediction Models Using the Matched Molecular Pair Kernel

Activity cliffs (ACs) are formed by two structurally similar compounds with a large difference in potency. Accurate AC prediction is expected to help researchers’ decisions in the early stages of drug discovery. Previously, predictive models based on matched molecular pair (MMP) cliffs have been proposed. However, the proposed methods face a challenge of interpretability due to the black-box character of the predictive models. In this study, we developed interpretable MMP fingerprints and modified a model-specific interpretation approach for models based on a support vector machine (SVM) and MMP kernel. We compared important features highlighted by this SVM-based interpretation approach and the SHapley Additive exPlanations (SHAP) as a major model-independent approach. The model-specific approach could capture the difference between AC and non-AC, while SHAP assigned high weights to the features not present in the test instances. For specific MMPs, the feature weights mapped by the SVM-based interpretation method were in agreement with the previously confirmed binding knowledge from X-ray co-crystal structures, indicating that this method is able to interpret the AC prediction model in a chemically intuitive manner.     

Aryloxy and benzyloxy compounds of zirconium: Synthesis, structural characterization and studies on solvent-free ring-opening polymerization of ?-caprolactone and δ-valerolactone

A series of aryloxy compounds and benzyloxy derivatives of Zr(IV) were synthesized in good yields and purity, employing the alcoholysis route. They were completely characterized with different spectroscopic techniques and single crystal X-ray diffraction. A high degree of fluxional behavior of these compounds was understood through variable-temperature NMR studies. X-ray diffraction studies prove that these compounds exist as dimers in the solid state. They are potent catalysts for the ring-opening polymerization of ?-caprolactone (CL) and δ-valerolactone (VL) resulting in high polymers with good number average molecular weights (M_n) and molecular weight distributions (MWDs). The degree of control in these polymerizations was found to be superior with the zirconium compounds when compared with the titanium analogues. The rate of polymerization was found to be slower for the zirconium compounds, as realized from kinetic studies. Analysis of the MALDI-TOF and ¹H NMR spectra of low molecular weight oligomers of CL synthesized using these compounds reflect that these polymerizations proceed by the activated monomer mechanism.     

25‐Hydroxyvitamin D3 Synthesis by Enzymatic Steroid Side‐Chain Hydroxylation with Water

The hydroxylation of vitamin D₃ (VD₃, cholecalciferol) side chains to give 25‐hydroxyvitamin D₃ (25OHVD₃) is a crucial reaction in the formation of the circulating and biologically active forms of VD₃. It is usually catalyzed by cytochrome P450 monooxygenases that depend on complex electron donor systems. Cell‐free extracts and a purified Mo enzyme from a bacterium anaerobically grown with cholesterol were employed for the regioselective, ferricyanide‐dependent hydroxylation of VD₃ and proVD₃ (7‐dehydrocholesterol) into the corresponding tertiary alcohols with greater than 99 % yield. Hydroxylation of VD₃ strictly depends on a cyclodextrin‐assisted isomerization of VD₃ into preVD₃, the actual enzymatic substrate. This facile and robust method developed for 25OHVD₃ synthesis is a novel example for the concept of substrate‐engineered catalysis and offers an attractive alternative to chemical or O₂ /electron‐donor‐dependent enzymatic procedures.     

ADME evaluation in drug discovery. 6. Can oral bioavailability in humans be effectively predicted by simple molecular property-based rules?

A critically evaluated database of human oral bioavailability for 768 chemical compounds is described in this study (http://modem.ucsd.edu/adme), which provides the scientific community a publicly available and reliable source for developing predictive models of human oral bioavailability. The correlations between several important molecular properties and human oral bioavailability were investigated and compared with an earlier report by analyzing the rat oral bioavailability data (J. Med. Chem. 2002, 45, 2615). We showed that the percentages of compounds meeting the criteria based on molecular properties does not distinguish compounds with poor oral bioavailability from those with acceptable values, which may suggest that no simple rule based on molecular properties can be used as general filters to predict oral bioavailability with high confidence. A data set of intestinal absorption was also examined and compared with that of oral bioavailability. The performance of these rules based on molecular properties in the prediction of intestinal absorption is obviously much better than that of oral bioavailability in term of false positive rate, and, therefore, the applications of the "rule-based" approaches on the prediction of human bioavailability should be very cautious.     

Pharmacophore Model Generation Based on Pyrrolidine- and Butane-derived CCR5 Antagonists

A three-dimensional pharmacophore model was developed for a considerable number of pyrrolidine-based and butane-based chemokine (C-C motif) receptor 5 (CCR5) antagonists, which can block the entry of human immunodeficiency virus type 1 (HIV-1) by inhibiting the interaction of HIV-1 envelope protein and CCR5. The pharmacophore model was generated using a training set consisting of 25 carefully selected antagonists with the diverse molecular architecture and bioactivity, as required by the Catalyst/HypoGen program. The activity of the training set molecules expressed in IC₅₀ (half-inhibitory concentration) covered from 0.06 to 10000 nmol·L^–1. The most predictive pharmacophore model (Hypo 1), consisting of two positive ionizable points and three hydrophobic groups, had a correlation of 0.924 and a root mean square of 1.068, and a cost difference of 63.67 bits between the null cost and the total cost. The model was applied in predicting the activity of 74 compounds as a test set. The results indicated that the model was able to provide clear guidelines and accurate activity prediction for novel antagonist design.     

Deuterium diffusion in vanadium deuterides (VDx; 0.4?x?0.6) studied by H NMR

Sites and dynamics of deuterium in VD_0.50 as well as in VD_0.40 and VD_0.57 have been studied by means of X-ray diffraction, differential scanning calorimetry and ²H NMR. VD_0.40 consists of the body-centered-cubic (bcc) α_D and body-centered-tetragonal (bct) β_D phases at room temperature and only the octahedral (O) sites are occupied in the β_D phase. On the other hand, in VD_0.50, consisting of the bctβ_D phase at room temperature, the O site is dominantly occupied with a small occupancy of the tetrahedral (T) sites in the β_D phase. VD_0.57 is composed of the bccα′_D (a high-deuterium-concentration phase of the α_D phase) and bctβ_D phases at room temperature, and the occupancy of the T sites is observed. The mean activation energy (E_D) of the deuterium diffusion is much larger for the O sites than for the T sites, and the diffusion is also much slower in the O sites. The mean E_D values for the O sites are almost the same in the β_D phase of VD_0.40 and VD_0.50, although the distribution of the mean residence time is much larger in VD_0.50 than in VD_0.40.     

New applications of the VDmax approach to substantiation of preselected sterilization doses

The VD_max approach for substantiation of 25 kGy has been in use for more than 10 years. VD_max methods are included in ISO 11137-2:2006 and AAMI TIR33:2005. ISO Technical Specification, 13004, is under development that will include sterilization doses from 15 to 35 kGy.For substantiation of a sterilization dose for very low bioburden products, less than or equal to 0.3, values of VD_max have now been derived and tabulated for a sterilization dose of 12.5 kGy.Products have been encountered that have both low bioburden and a relatively low maximum dose. In several situations, existing tabulated VD_max values could not be effectively used; in one such situation, the average bioburden was too high to substantiate a 15 kGy sterilization dose and the use of a 17.5 kGy sterilization dose was not practicable due to the likelihood of exceeding the product's maximum acceptable dose. For this product, values of VD_max were derived and tabulated for substantiation of a 16.1 kGy sterilization dose.Values of VD_max have been derived and tabulated for the substantiation of sterilization doses linked to a sterility assurance level (SAL) of 10^?3. To offer a potential alternative to aseptic processing, the notion of using an “aseptic processing equivalent dose”, 10^?4 SAL, has been investigated along with the use of alternate model populations for calculation of VD_max values.