Fig 1.
Comparisons of clustering results relative to two different measures of similarity for the InhA enzyme MD trajectory.
In each case, the k-means clustering algorithm with the number of clusters of eight was used. The difference of the cluster dispersion between the traditional method and our methodology is evidenced by using the pairwise RMSD distance and the properties from the substrate-binding cavity in the scatter plots (a) and (b), respectively.
Fig 2.
Substrate-binding cavity of the InhA enzyme (PDB ID: 1BVR) identified by the CASTp software tool.
On the left, the substrate-binding cavity of the 1BVR structure represented by molecular surface and colored by atom types (carbon and hydrogen: light grey; nitrogen: blue; oxygen: red; sulphur: yellow). The projection displays all residues from the binding pocket in stick representation.
Table 1.
Fragment of the Cavity Attributes data set used for clustering the MD trajectory.
The first line labels the substrate-binding cavity features. The number below each residue indicates the maximum number of heavy atoms it can hold.
Fig 3.
3D structure of the 20 ligands used in docking experiments.
Each ligand is colored by atom type (carbon and hydrogen: light grey; nitrogen: blue; oxygen: red; chloro: green; phosphorus: orange; sulphur: yellow) and displays its name and PDB identification (PDB ID). The dashed circle represents the rotatable bounds selected by AutoDockTools 1.5.6. Ligand Abbreviations: TCL: triclosan; 665:(3S) N-(3-bromophenyl)-1-cyclohexyl-5-oxopyrrolidine- 3-carboxamide; 566:(3S)-1-cyclohexyl-5-oxo-N-phenyl pyrrolidine-3-carboxamide; 8PC: 2-(2,4-dichloro-phenoxy)-5-(pyridin-2-ylmethyl)phenol; JPJ: 2-(2,4-dichlorophenoxy)-5-(2-phenylethyl)phenol; JPL: 5-(cyclohexa-1,5-dien-1-ylmethyl)-2-(2,4-dichlorophenoxy) phenol; JPM: 5-benzyl-2-(2,4-dichloropheno-xy)phenol; 468:(3S)-N-(3-chloro-2-methylphenyl)-1-cyclohe xyl-5-oxopyrrolidine-3-carboxamide; 641:(3S)-1-cyclohexyl-N-(3,5-dichlorophenyl)-5-oxopyrrolidine-3-carboxamide; 744: (3S)-N-(5-chloro-2-me-thylphenyl)-1-cyclohexyl-5-oxopyrrolidine-3-carboxamide; INH-NAD: Isoniazid + NADH coenzyme; 5PP: 5-pentyl-2-phenoxyphenol; 8PS: 5-octyl-2-phenoxyphenol; TCU: 5-hexyl-2-(2-methylphenoxy)- phenol; PTH-NAD: Prothionamide + NADH coenzyme; THT: trans-2-hexadecenoyl-(n-acetyl-cys-teamine)- thioester 4PI: N-4-methylbenzoyl-4-benzylpiperidine; GEQ: 5-{[4-9H-fluoren-9-YL)pipera- zin-1-YL]carbonyl}-1H-indole.
Table 2.
Summary of docking experiments performed to analyze the clustering results.
The sixth column highlights the cross-docking experiments that showed RMSD values above 2.0 Å for the best FEB values (kcal/mol). For comparison, the third and fourth column indicates docking results (FEB and its corresponding RMSD) obtained by reproducing the original pose of the ligand in its crystal structure.
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 (a) and (b), respectively. The black points identify the optimal partitioning solutions.
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 as a function of the number of clusters for, UPGMA, WPGMA, Complete and Ward’s methods are showed in the graphs (a), (b), (c) and (d), respectively. The black points identify optimal partitioning solutions.
Table 3.
Statistical evaluations for the optimal partitioning solutions obtained from the best partitions (lowest SQD value) of every clustering method.
Third column indicates the number of medoids used in the statistical assessments. Average, standard deviation and variance were calculated for each set of medoids based on predicted FEB values. Last row indicates the statistical values for the MD’s full trajectory.
Fig 6.
The ensemble of representative MD conformations generated from the best partitioning solution.
The boxplot represents the trends in docking result changes in median FEB per data set and per ligand. The data sets are represented by different colors, where, pink, yellow, and green are the data sets used in the clustering experiments and blue is the original simulation.