7.2 CaverDock cannot compute the upper-bound trajectory 12
from moving through the tunnel. The receptor geometry needs to be fixed by adding
flexibility to side-chain residues, or using a different snapshot of the receptor.
• The ligand is forced to move against the active site bottom. This problem arises from a
geometrical analysis of the tunnel where the geometrical approximation of the tunnel is
too deep. An easy solution is to use the --dock like parameter to navigate CaverDock
where to start with the tunnel analysis.
• The ligand is too complex to be successfully docked. CaverDock can be re-executed
with higher exhaustiveness, or a higher number of parallel workers.
• The computation fails due to the stochastic nature of CaverDock. In this case, starting
CaverDock once again should solve the problem.
7.2 CaverDock cannot compute the upper-bound trajectory
When an upper-bound trajectory cannot be computed, the user should inspect the lower-
bound trajectory first. When the lower-bound trajectory already contains a high energetic
barrier, the tunnel or the receptor geometry needs to be modified (Section 6.2). If it is not
possible to improve the lower-bound trajectory, the selected ligand is likely not able to pass
through the tunnel.
When the lower-bound trajectory does not contain any significant barrier and the upper-
bound trajectory is still not computed, CaverDock is either not able to analyze the trajectory
because of a high ligand complexity, or there is a bottleneck not detected by the lower bound
trajectory. This type of bottlenecks can be found by inspecting the lower-bound trajectory
in the vicinity of the disc, where the computation of the upper-bound trajectory has failed
3
.
There should be visible non-contiguities in the lower-bound trajectory, which overpass the
bottleneck (e.g. ligand flip, as shown in Fifure 3). The bottleneck needs to be manually
identified and fixed (e.g. by using flexible side-chain residues).
8 FAQ
8.1 Is CaverDock execution deterministic?
In the default settings, it is not deterministic (so re-executing CaverDock may result in dif-
ferent trajectories). There is, however, a way how to make CaverDock deterministic. It must
be executed in two processes only (using mpirun -np 2) and the random number generator
must be set to a constant seed (using --seed x, where x is any number, which must be the
same for all deterministic executions). Please note than when CaverDock is executed only for
single docking (possibly with constraints), setting a constant seed is sufficient.
8.2 Why does CaverDock return two types of energy?
The CaverDock computation of an upper-bound trajectory is driven by a heuristics. It means
that CaverDock cannot guarantee that the contiguous trajectory is optimal. Thus, CaverDock
is computing also a lower-bound trajectory: a scenario, which can be unrealistically optimistic
3
CaverDock reports a number of the last disc, for which the upper-bound trajectory was computed. The
number of snapshot in the lower-bound trajectory is equal to the number of the disc.