This page shows a Hierarchical Graph representation of Pharmacophore Models generated from a molecular dynamics simulation of the 4no7 crystallographic structure.
Brief view explanation
The view is best working with the Google chrome internet browser.
Each node of the graph represent a unique pharmacophore model. Mousing over a node will display its feature composition.
Clicking on a node will highlight all nodes that are in a subset or superset relationship with the current selection in terms of pharmacophore feature composition.
Their associated feature vector will be displayed on the top of the graph, as well as the MD frames in which corresponding pharmacophore models are observable.
Several nodes can be selected simultaneously to highlight their common features and the shortest path that link them. Clicking aside nodes will reset the current selection.
Certain nodes were not directly observed but were generated to obtain a unique graph regrouping all observed models. The generated nodes are orange, and the observed one are blue.
The size of the node is directly linked to the number of time this specific pharmacophore model was observed during the simulation.
Nodes Ids are decreasingly related to the node size, meaning the node 0 has the highest observed frequency.
The nodes are organized by feature number on the x axis and relative distance (MDS, Manhattan distance) on the y axis.
The variance of the projection is shown in the "Infos" table. The nodes were slightly moved to avoid overlapping.
All nodes corresponding to unique pharmacophore models that were observed less than 10 times during the simulation were removed, except for the node 176 that correspond to the pharmacophore model from the crystallographic structure.
The slider "Pharmacophore filter" allow the highlighting of single nodes based on their Id.
Clicking on a specific feature in the feature vector display below the slider allow the highlighting of nodes composed of this feature.
Mousing over the elements of the feature vector will display the labels of the unique pharmacophore features.
In this example, the label shows the type of pharmacophore interaction, the atoms in the ligand involved in the interaction and then the amino acid involved from the protein side.
A typical use of this view for the selection of a single pharmacophore for virtual screening purposes would be to consider the node with the highest frequency (node 0),
perform a virtual screening with one frame where this model is present and then add or remove features of this model following hierarchical links to achieve the best
ratio between accuracy and specificity for your project.