|BioPredict has developed a
suite of hypothesis-driven docking and ab-initio ligand design
programs for use in virtual screening. Careful comparison of
crystallized ligand-protein complexes of the target or related targets
can often identify one or more binding motifs to serve as hypotheses in
identifying new active compounds and in predicting their correct
placement in the active site. The software platform described distills these
hypotheses into geometrical and chemical constraints that are then
incorporated directly into the docking process. Current docking-based
screening methods that rely on energy criteria alone have a high failure
rate, both for identifying new active molecules and for predicting
correct binding modes1. Hypothesis-driven docking alleviates
this problem by augmenting energetic criteria with prior knowledge
incorporated directly in the docking process.
Constraints can include but are not limited
to: (i) hydrogen bonds and salt bridges between ligand atoms and
designated protein atoms, (ii) distance constraints from ligand atoms to
protein atoms or to geometrically placed points in the docking site, and
(iii) occupation of hydrophobic pockets. Formally a constraint can be
defined as a set of limits on any subset of interaction energy terms
describing the interaction of a ligand with any subset of protein atoms
in the active site.
The docking procedure involves breaking
the ligand into constituent fragments and recursively docking larger
fragment constructs starting from the largest fragment of the ligand.
The ab-initio procedure is similar but it selects its next
fragment and linker from a library. Many possible ligand poses are
retained at every stage of this procedure. Energetic optimization is
performed at each stage by adjusting translation, rotation, ligand bond
rotation, and side-chain rotations for selected protein side chains.
Constraints are imposed using the Method of Langrange Multipliers.
1 Docking on Trial, Peter
Kirkpatrick, Nature Reviews Drug Discovery 4: 813 (2005).