Control the Cost of a Gigadock Run

Context

Running the Gigadock Floe on billions of molecules generally incurs significant computational cost. Typically, a Gigadock run using 2–3 billion molecules costs approximately $30K, although there is significant variance depending on the characteristics of the target and molecules being docked. Costs could run from $10K to more than $100K.

This describes what factors affect the time and cost of the Gigadock Floe and what changes an end user can make to adjust those factors.

Receptor / Design Unit

Size of the Active Site

The size of the active site is strongly correlated with the time and cost of docking; reducing the size will reduce the docking time and cost. The most relevant measure of active site size is the volume of the outer contour, which is a volume complementing the active site and which all heavy atoms of docked poses must fit within. The outer contour can be visualized in Orion. To measure or adjust the outer contour volume, see the UI Guide.

Note

The maximum recommended output contour volume is 1000 Å³.

The shape of the outer contour is used to filter the poses of a molecule whose conformers have been exhaustively rotated and translated within the active site. After this filtering, the poses are all scored, which consumes the vast majority of the computational cost relative to the outer contour filter. Increasing the outer contour volume increases number of poses that need to be scored, thus increasing the time and cost of the docking. This effect is particularly severe in large active sites where the outer contour encloses large amount of solvent space not particularly near the protein surface. In this case, large numbers of poses that interact mostly with solvent pass the outer contour and get scored.

Enabling an inner contour should reduce computational cost slightly. However, if the inner contour size is nearly as large as the outer contour, there will be little to no savings.

Docking Constraints

Adding one or more docking constraints to the receptor/design unit will reduce the time and cost of docking. The more constraints added, the more the docking time and cost will be reduced, although there is a diminishing effect of additional constraints. Constraints are visible in the 3D Viewer, where they can be added, deleted, or adjusted.

Docking constraints filter out poses that do not interact with the active site specified by the constraint(s). This happens after the docking molecule’s conformers are exhaustively rotated and translated with the site and the resulting poses filtered by the outer contour (see previous section), but before the poses are scored. The vast majority of the computational time goes to the scoring, so filtering out more poses by adding constraints improves the speed of the docking. The more restrictive the constraint(s), the more poses are filtered before scoring and hence the faster the docking speed.

Docking Mode

The selected docking mode also affects the time and cost of a Gigadock run. The Gigadock Floe has three docking modes (see the Dock Method parameter under Options in the Job Form of the Gigadock Floe). The relative speed of these three methods is listed from slowest (highest cost) to fastest (lowest cost):

Fred
Hybrid
Fast-Fred

The default method is ‘Fred’.

Docking Molecules

Number of Molecules

The overall time and cost to run a Gigadock job is proportional to the number of molecules docked. Docking a smaller set of molecules will generally be faster.

Note

The size and number of conformers of docked molecules also influences how fast they dock. It is possible for a small set of large, floppy molecules to dock faster than a larger set of small, rigid molecules.

Filtering Molecules

The number of molecules to dock can be reduced (and hence the cost of a Gigadock run lowered) if there are known molecular properties the docked molecules must have. The Filter Collection Floe can be used to filter an existing docking collection based on a variety of molecular properties. Alternatively, if the docking collection is being created from scratch from a SMILES or other file, the same set of filtering options is available in the Prepare Giga Collections Floe.