Extended Data Fig. 3: Cluster analysis of the 45 input structures for free energy perturbation simulations.
From: Structural insights into tecovirimat antiviral activity and poxvirus resistance

Analysis of putative binding poses after equilibrating the 15 selected poses (x-label, first number) by three independent MD replicas (x-label, second number, see Methods). To reveal reoccurring protein-ligand interaction networks, a cluster analysis was performed. The x-axis represents the pose identification, and the y-axis indicates the distance between each pose in terms of their protein-ligand interaction network. The lowest energy pose of each cluster is highlighted, with its protein–ligand interaction profile and corresponding absolute binding free energy. The analysis demonstrates (together with Supplementary Table 4) that different protein-ligand interaction networks yield similar absolute binding free energies. The pose with the strongest binding affinity of -25.6 kcal/mol was used for further refinement against the crystallographic data, as reported in the PDB file. The Python library ProLIF73 was used to generate the protein-ligand interaction network.