[Product Releases]
|
|
|
|
|
[Blog]
|
|
Most recent post
|
[News]
|
|
Can we trust docking results? Sept 2010 IBM Systems and Technology Group releases a white paper with eHiTS and Cell
Oct 2008
EPA's ToxCastTM project will use SimBioSys' eHiTS as docking engine
Nov, 2007
|
[Events]
|
| 240th ACS
Aug 22-26, 2010 Boston, MA, USA
booth #945
see >> more
|
|
|
|
|
Force field, energy calculation problems
-
Exact atom-atom calculations (e.g. MM):
-
Too slow for protein-ligand interactions to be performed during a (flexible)
docking study or de novo design, where millions of structures might
have to be evaluated against the same receptor.
-
Grid-based techniques:
-
Lose-lose accuracy-speed trade-off: manageable resolution yields poor approximations,
better results require too large grid, which is too slow to precompute.
-
Large memory requirement for good resolution (e.g. 0.1A): the required
memory for resolution giving good results is in the range of gigabytes,
which is too much to handle in physical memory.
-
Precompute values at irregular point sampling (high resolution at interaction
surface only):
-
Searching complexity for queries
-
Discrete results, bad for optimization (not suitable for interpolation)
Solution: Property Fields
Our
solution to these problems is a new patent pending data structure and accessing
algorithm. We call it a Property Field.
Features
-
Supported interaction field potentials include: Electrostatic, H-bonding, Hydrophobicity,
van der Waals (Lennard Jones) and steric proximity potentials
can be queried with vectorial information where applicable (e.g. optimal
hydrogen bonding vector).
-
The implementation allows dynamic localized update of the data, i.e.
if the receptor structure is changed, the Property Field data
can be updated without full recalculation - changing the affected values only.
-
The technique is applicable to the calculation of any standard force-field
Benefits
-
Smoothly changing values given for any 3D location, which is suitable for
steepest decent or other energy optimisation techniques, that require
differentiable functions.
-
Low
memory requirement: comparable to a coarse (e.g. 1.5A edge) grid representation.
-
High accuracy: exceeds the accuracy of a fine (e.g. 0.1A edge) grid representation,
can closely approach the exact atom-by-atom calculation values.
-
Good performance: is provided by constant access time of the data, in contrast
to searches required for irregular spacing grids (e.g. octrees) that could
share the low memory and high accuracy features of this system.
Summary
This system overcomes the quality-speed compromise of all other techniques
(precise atom-by-atom calculations, regular and irregular grids).
|
|
|