OverviewΒΆ
EON calculates the electrostatic similarity between two small
molecules in the form of an Electrostatic Tanimoto (ET) score. EON supports
two descriptions of electrostatics: via charge density (by default)
and via potential (-potential
true) Given a query molecule and a set of
interesting molecules, EON will calculate the Electrostatic
Tanimoto between each database molecule and the query. Note, that starting
from version 3.0, EON does perform an overlay of the input structures using molecular
shape and charge density similarity optimization. Since
electrostatics calculations require high quality partial charges, EON will
calculate new partial charges for the input structures using
MMFF94. If the user provides an input file that contains
structures with higher-quality partial charges, EON can use them as
well (-charges
existing).
EON is also dependent on pKa state and formal charges as these have a significant impact on electrostatics. EON now has the ability to adjust both the query and database molecule to a neutral pH model. This feature is on by default, but can be turned off by using appropriate command line flags.
When EON is run in -potential
true mode, the best shape and charge density hits
are rescored with shape and potential similarity. In doing so the command line flag -num_top_scores
(by default equal to 5) controls how many best hits per molecule are being re-scored.
Part of understanding EON results is visualization of the electrostatic grids used in the calculation.
Although off by default, when writing EON results to a binary (OEB) file, ET grids can be attached to each molecule
and visualized using the EON View mode in VIDA. When running with -writegrid
true,
for -potential
false the charge density grid will be outputted and for -potential
true the
electrostatic PB potential grid will be added to results.
Since EON calculations can be time-consuming (approximately 1 molecule per second per CPU for
the -potential
true case), EON can use the same distributed computing technology,
Open MPI, that ROCS uses to help distribute the workload across a cluster of machines.