Chemistry Reference
In-Depth Information
network. The number of neurodes and weights are used to modify the initial input
signal propagation. In the output layer of artificial neurons, analysis of the original
input signal is ultimately encoded [346]. For scoring evaluations the strengths of
the connection between neurodes can be varied until the network could reliably
predict binding affinities when given descriptors of ligand-receptor complexes.
A specific receptor tailored composite scoring function has been designed to
optimize ranking of known inhibitors can potentially enhance virtual-screening
performance. The best protocol to use in a virtual-screening project is highly
dependent on the target receptor studied.
Recent work suggest improving the scoring of protein-ligand binding affinity by
including the effects of structural water and electronic polarization. The authors
use minimization, MD, MM/PBSA and their incorporated specific charge model
(PPC) to obtain correct ranking of binding poses in which bridging water
molecules and electronic polarization play an important role [553].
Recent studies report optimization of molecular docking scores with 'support
vector rank regression' (SVM), a statistical learning strategy that in principle
represents a major advance in machine leaning methodology [554]. SVM uses
training examples to construct a decision hyperplane in a high-dimensional feature
space. The decision margin is maximized in order to minimize the generalization
error. It is often implemented as a classification algorithm that learns from
categorical examples to classify new instances.
It is also implemented as a regression algorithm that learns from continuously
varying examples to predict the target value of new instances. Using different
example datasets and training strategies the authors conclude that with additional
features for comparative fitness between computed binding conformations, the
algorithm holds the potential to create a new category of more accurate integrative
docking scores.
ENTROPIC CONTRIBUTIONS
A long-standing challenge is the rigorous evaluation of entropic fluctuations
during macromolecular associations in molecular modeling. Effectively, ligand
