Chemistry Reference
In-Depth Information
proteins. In the mathematical GRIND approach (Grid-Independent descriptors)
descriptors are calculated independently of the orientation in space. The most
important grid interactions as a function of distance are represented. This
approach represent a class of alignment-independent three-dimensional molecular
descriptors which are important for biological property description. Molecular
interactions fields are simplified and encoded using autocorrelation transforms.
The resultant molecular descriptors yields correlograms (graphical diagrams) used
in chemometric analyses.
Using 'grid modes', analytic MIF expressions can be sampled over space
surrounding the molecules at regular intervals. These method can yield hundreds
or thousands of data points including 'hot spots' (regions of space with the most
favorable energies (favorable binding sites). Algorithms can also be used to
extract sets of hot spots from MIFs [723]. The node of the MIF can be tagged by
the atom contributing most to the field energy. Energy cutoffs are used for node
prefiltering making it possible to discriminate between weak and relevant nodes.
Grid MIFs are also used with high-throughput virtual screening of proteins. The
binding site is first described and then fingerprints for ligand and proteins (FLAP)
methods are determined and used to encode and compare the information. GRID-
MIFs and GRID-FLAP methods are also used for identifying new targets. There
are diverse methods that use MIFs to identify protein structure with high
propensity for interaction with ligands [744].
3D molecular similarity measures used for matching groups of atoms
via
geometrical criteria can miss topologically different molecules that exhibit similar
biological effects. Two molecules can have similar biological properties with
different covalent structures and chemical formulae. Usage of molecular
electrostatic potential (MEP) is a protocol that avoids limitations of atomistic
models. Various field-based methods use 3d grids for comparing and sampling
MEPs. It is noteworthy, however, that compared to other 3d-LBVS techniques,
this can be computationally expensive. Global superposition can be misleading
because of pitfalls in defining initial structural superposition as well as best
overlaps in a specific region of interest [384-386].
