However, whether the membrane fusion is influenced by -carrageenan is not clear

D value less or equal to 2.0 from the crystallographic position of the ligand. Clustering analyses on data sets from the MD trajectory This section reports and compares the results obtained for clustering three different data sets from structural information of the FFR model. We applied the six clustering algorithms described in the Materials and Methods section. In this regard, we first executed the clustering algorithms for get LBH589 Cavity Attributes, Cavity RMSD, and Protein RMSD data sets varying the number of clusters from 10 to 200, and afterward we extracted the medoids from every generated partitioning. Solutions were evaluated based on statistical assessments in the predicted FEB values. We decided to start the clustering analyses from 10 since low k values shows poor level of scatter and, consequently are unable to reflect all possible movements of a 20 ns MD trajectory. In opposition, high numbers of clusters tend to represent better dispersion but we limit the cluster ranges up to 1% of all MD conformations since our findings show the best partitioning solution used cluster count less than 100. Our first set of experiments was performed with a number of clusters range from 2 to 1,000. However, we decided to decrease this range for two reasons: the time consuming taken for performing practical virtual screening of large database of ligands in an ensemble with 1,000 representative MD conformations; and the high level of accuracy achieved by using a representative ensemble with 200 MD conformations. To support the second reason above described, we analyzed and compared all clustering solutions taking into account the level of coverage reached by them in terms of dispersion and MD trajectory representativeness. The dispersions among the partitions generated from 10 to 200 clusters were analyzed by assessing the SQD values. The resulting SQD values by clustering method for Attribute, Cavity RMSD and Protein RMSD data sets are in S1, S2 and S3 15 / 25 An Approach for Clustering MD Trajectory Using Cavity-Based Features Fig 4. Comparative performance of partitioning clustering methods for the three data sets under study. Variations in the SQD values as a function of the number of clusters for k-means and k-medoids are showed in the graphs and, respectively. The black points identify the optimal partitioning solutions. doi:10.1371/journal.pone.0133172.g004 16 / 25 An Approach for Clustering MD Trajectory Using Cavity-Based Features Fig 5. Comparative performance of hierarchical agglomerative clustering methods for each of the three data sets generated from the conformations of the MD trajectory. The SQD values PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19748594 as a function of the number of clusters for, UPGMA, WPGMA, Complete and Ward’s methods are showed in the graphs,, and, respectively. The black points identify optimal partitioning solutions. doi:10.1371/journal.pone.0133172.g005 17 / 25 An Approach for Clustering MD Trajectory Using Cavity-Based Features doi:10.1371/journal.pone.0133172.t003 count, 18 / 25 An Approach for Clustering MD Trajectory Using Cavity-Based Features achieving the similar central tendency undertakes for the entire ensemble of MD conformations. As expected, the partitions from Cavity Attributes data set appear to accurately determine crucial changes that occur in the substrate-binding cavity PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19748686 of the MD conformations under study. Fig 5 evidences this statement by drawing the Cavity Attributes analyses with lower SQD values and The representative ensemble of MD rece