Methods Patients Thirty-two adults receiving living donor kidney allografts were studied

ting partitions for evaluating binding conformations of different drug candidates to the enzyme under study. In this regard, 20 compounds experimentally tested were used in the cross-docking experiments against the FFR model to predict their lowest energy bound conformation. This range represents the different sizes and structures of the ligands that are bound in such structures, being a good sample of structures to dock in our MD trajectory, which has its binding cavity volume ranges from 419.4 3 to 2,032.8 3. The docking experiments were performed on FReMI, which is a MEK 162 price powerful middleware developed to execute molecular docking simulations of FFR models in multiprocessing machines using the AutoDock4.2.5.1 software. We set up AutoDock4.2 to run the Lamarckian Genetic algorithm on three Intel Core i7-2600 3.4 GHz computers with 12 GB RAM, using as operating system the Linux Ubuntu version 13.04. The number of docking runs was set up to 25 since we are working with flexible structures. The box dimensions for grid parameters were PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19747723 tailored according to the structure of each ligand. We defined the atom types of AutoDock, add the Gasteiger charges and merge the non-polar atoms for each snapshot of the FFR model. All these configurations were determined before running the docking experiments and the input AutoDock files were generated during the experiments by FReMI based on AutoDockTools 1.5.6 parameters. The energy evaluation and the number of generations were configured as 300,000 and 27,000, respectively. All ligands were treated as flexible in AutoDock and their rotatable bounds are highlighted in Fig 3. In order to preserve the environmental conditions available to the substrate and ligands, the NADH coenzyme was considered as part of the protein receptor. Conversely, we removed the NADH coenzyme from all snapshots of the FFR model when we performed the experiments with adducts, since they already have the coenzyme as part of their structures. To obtain ensembles of representative MD conformations from the clustering algorithms used we assessed the dispersion of the resulting partitions based on the predicted FEB values. The dispersion corresponds to the first, second and third quartile values that were calculated to compare the level of convergence between the resulting partitions and the MD’s full trajectory. The first step in this process was to extract the medoids from all partitioning and obtained an ensemble of medoids for every partitioning. In this study, we analyzed partitions from three different data sets and six different clustering methods. Even though we are using FEB predicted values to select and validate the optimal partitions, we believe that our methodology can provide an effective strategy for improving clustering MD trajectory approaches. Further, we expect that the best set of medoids can also be able to reproduce a fairly good level of distinct binding models for different ligands–especially when the Cavity Attribute data set is used as input for the clustering algorithm. Results One of the major challenges in performing docking experiments of FFR models is the computational demand to screen large databases of small compounds and extract potential binders. According to Amaro et al., performing virtual screening experiments in the full set PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19748686 of structures is computationally intractable and likely unnecessary. Zhong et al. and Cheng et al. show that using a minimal representative ensemble of MD conformations is a pr