In the absence of target structure, LBDND approach can be
used by gathering constraints from the known active and inactive
ligands of primary target [25]. Furthermore, ligands contain the
hidden information about the target binding pocket that can be
revealed by the superimposition of 3D ligand structures over each
other and generating a pseudoreceptor binding pocket image
[26]. Finally, 3D QSAR model based on selective and common
pharmacophores can be generated to quickly facilitate the search of
ligands [27].
There are many methods reported to explore the receptor
binding site for SBDND such as “rule based approach,” “grid-
based methodology,” and flooding of protein binding sites with
fragments [28–33].2.2 Design
Constraint: Secondary
Target
A drug should not only be an active molecule but should also be
biologically active. Possession of suitable drug-like properties such
as absorption, distribution, metabolism, excretion, and toxicity
(ADMET properties) makes an active molecule to be biologically
active. Constrains other than those that define binding affinity are
considered secondary constraints. The overall de novo scoring is
adjusted accordingly based on the weightage of secondary con-
straints. The prediction models can be as simple asLipinski’s rule
of fiveor as complex as deriving the bioavailability based on cell line
studies [34, 35].2.3 Scoring Function Scoring functions are used (1) to control and direct the ligand
design process and (2) to estimate the binding affinity and rank
the generated novel molecules. The scoring function must be accu-
rate enough to avoid false positive and false negative selections from
the chemical space and fast enough to quickly explore the vast
chemical space at the same time. In SBDND, scoring functions
take advantage of available 3D coordinate information about the
target molecule and are very similar to those used for molecular
docking process. All the quality assessment scoring functions are
approximations and their algorithms can be classified into three
categories (1) explicit force-field methods, (2) empirical scoring
functions, and (3) knowledge based scoring functions. Explicit
force-field methods are computationally more demanding but are
expected to produce more accurate results. Empirical scoring func-
tions are based on the weightage sum of the receptor–ligand inter-
action types, where few interactions are favored over other and
unfavorable interactions leads to penalty. The interaction types
include the H-bonds, electrostatic interactions, hydrophobic inter-
actions, and also inclusion of aromatic interactions in recent years.
Empirical methods provide the computational speed without sacri-
ficing much on the accuracy, and are specifically effective in ranking
the molecules. Knowledge based scoring functions are based on the
statistical analysis between ligand–receptor complex structures.
Fragment-Based Ligand Designing 127