ArticlePDF Available

Abstract and Figures

The discovery of novel biologically active small molecules is now a technologically and economically viable proposition for academic and small biotechnology laboratories wishing to build on their biological research into target proteins. Such small molecules may be useful reagents for further biological research or may form the basis for early-stage drug discovery. The availability of specialized virtual screening software to filter large molecular libraries into manageable numbers of compounds for biological assays is the driving force for finding novel ligands. The main focus of this chapter is the basis and use of molecular field methods to assess the interactions that may be made by small molecules. Molecular field based measures of capability and similarity of interaction may be used to discover novel ligands and expand ligand series for potential use as future therapies.
Content may be subject to copyright.
Chapter 18
Molecular Fields in Ligand Discovery
Paul J. Gane and A.W. Edith Chan
Abstract
The discovery of novel biologically active small molecules is now a technologically and economically viable
proposition for academic and small biotechnology laboratories wishing to build on their biological research
into target proteins. Such small molecules may be useful reagents for further biological research or may
form the basis for early-stage drug discovery. The availability of specialized virtual screening software to
filter large molecular libraries into manageable numbers of compounds for biological assays is the driving
force for finding novel ligands. The main focus of this chapter is the basis and use of molecular field methods
to assess the interactions that may be made by small molecules. Molecular field based measures of capability
and similarity of interaction may be used to discover novel ligands and expand ligand series for potential use
as future therapies.
Key words Intermolecular interactions, Drug discovery, Molecular fields, Virtual screening, Bioisos-
teres, Scaffold hopping
1 Introduction
I would like our field to be effective, one that contributes as much as possible
to the most important industry on earth—the discovery of these amazing
small molecules with their potential for dramatic effects on health and well-
being. Anthony Nicholls (1).
The search for new “amazing molecules” is the science and art of
drug discovery. Why art? Because at each stage within the drug
development pipeline decisions are made by individuals with a
personal bias based upon years of experience; this subjectivity, call
it art, is not easy to quantify even less, codify. However, the greater
our knowledge of science the better these decisions become and in
modern drug discovery a good understanding of the molecular
interactions which occur between a drug and its target molecule
can prove crucial for success.
Mark A. Williams and Tina Daviter (eds.), Protein-Ligand Interactions: Methods and Applications, Methods in Molecular Biology,
vol. 1008, DOI 10.1007/978-1-62703-398-5_18, #Springer Science+Business Media New York 2013
479
In most cases these intermolecular interactions are non-
covalent and weak, much weaker than the bonds between their
constituent atoms. It is this very weakness which provides a low
energy barrier for fast dynamic reactions to occur allowing life as we
know it to exist. Our understanding of these interactions is still
incomplete and our algorithms are based on approximations and
assumptions but despite this, computational chemistry or molecu-
lar modelling in general has become the leading tool in drug
discovery.
There are many excellent textbooks and articles (25) describ-
ing the vast range of computational methods available including
chapters within this volume; in an attempt to avoid repetition, only
a brief outline of intermolecular interactions will be given followed
by a more detailed description of one particular approach which
strives to visualize intermolecular interactions from a ligand mole-
cule’s point of view.
1.1 Intermolecular
Interactions
Let us step back for a moment and try to recall exactly what those
intermolecular interactions are. The potential energy curve, often
called a Lennard-Jones potential (Fig. 1), occurs in many introduc-
tory texts, and shows the effect of a long range attractive force
bringing the molecules together until they reach a minimum energy
at a separation r
m
(the force being zero at this point), and that as the
molecules are pushed closer together a repulsive force rapidly rises
to oppose the attraction (N.B. that the energy is still negative until
the collision diameter (σ) is reached). The total intermolecular
energy is a combination of the attraction and repulsion curves
shown as dot-dash lines (Fig. 1). Note that here the molecules
themselves are assumed to be spherical so there are no orientation
parameters to consider but in real, flexible molecules this simple
energy curve becomes a complex potential energy surface.
Fig. 1 The Lennard-Jones potential. Potential energy, U(r), and force, F(r),
versus inter-atomic/molecular distance (r) for neutral particles
480 Paul J. Gane and A.W. Edith Chan
The explanation of the multiple interactions which contribute
to the complete intermolecular potential energy surface was not
possible until the development of quantum mechanics. The attrac-
tive forces are a result of both classical electrostatics and quantum
effects, whereas the very short-range repulsive force is largely due to
the Pauli exclusion principle which disallows full electron shells
from overlapping. The attractive forces can be divided into a num-
ber of components:
lCharge–charge (Coulombic) interactions, if both molecules
carry a net charge.
lDipole–dipole (or multipole) interactions (Keesom forces), if
the molecules are polarized, i.e., they have a time-averaged
asymmetric distribution of charge.
lA charge or dipole (or multipole) in one molecule can distort
the electron cloud of a neighboring neutral molecule causing it
to become temporarily polarized, the two molecules are then
attracted to interact with the each other (Debye force). This is
known as induction.
lEven if the molecules do not carry a charge and are not polar-
ized there is a still a substantial, but short-range, interaction
between all molecules, this is a result of quantum fluctuations in
the electron orbitals which can induce correlated fluctuations in
neighboring molecules causing them attract via so called disper-
sion energy (London forces).
Coulombic interactions, i.e., those between charges (mono-
poles), weaken only slowly over distance (r) as a function of 1/r,
so they are very long range. Dipole–dipole is reduced more rapidly
by a factor of 1/r
3
. So polar molecules first interact by virtue of the
electrostatic forces which rapidly get stronger as the molecules
approach each other, as the distance reduces to around 6 the
short-range dispersion forces (1/r
6
) begin to add significantly to
the overall binding energy but if they get closer still within 2 or so
the large repulsion (1/r
12
) counteracts the attraction leading to the
equilibrium distance or van der Waals radius, identified as r
m
on
the Lennard-Jones curve. Another term often used to describe the
collective effect of the non-Coulomb terms is the “van der Waals
interactions,” which represents an aggregate of Keesom, Debye,
and London forces.
The size of the contributions of each component to the total
force or potential depends upon the chemical nature of the inter-
acting molecules. For example, water molecules interact mainly by
hydrogen bonds which can be considered as largely electrostatic/
dipole–dipole interactions; in fact each water molecule usually par-
ticipates in 3–4 hydrogen bonds, and this dominates their intermo-
lecular interactions and explains many physicochemical properties
of water. On the other hand, dispersion energy is the major
Molecular Fields in Ligand Discovery 481
contributor to HCl interactions, even though it is also a polar
molecule (6). The reason is twofold. First, chlorine is more polar-
izable than oxygen and as polarizability increases dispersion
becomes stronger. Second, the larger atomic volume of chlorine
reduces the ability of its lone pairs to form strong hydrogen bonds.
In addition to electrostatic, induction, dispersion and repulsion
terms some non-covalent interactions (e.g., hydrogen bonds or
halogen bonds) may have a significant quantum mechanical contri-
bution due to partial sharing of electrons between atoms (7).
However, such quantum mechanical effects are not usually explic-
itly modelled, but their effect is approximately included via some
modification of the parameters describing the interactions of the
groups involved.
We consider some of the more specific interactions which exist
between a small molecule (drug) and its protein target in the next
section. The reader is directed to the outstanding review by Bis-
santz (8) which describes these interactions from the perspective of
a medicinal chemist. Chan et al. (9) have analyzed the Protein Data
Bank (PDB) (10) to identify the most likely interacting small
molecular fragments for a given amino acid, which is both instruc-
tive and provides a basis for assessing the relative importance of
these interactions.
1.2 Hydrogen Bonds,
Salt Bridges, and Weak
Hydrogen Bonds
A “classical” hydrogen bond (H-bond) D–H:A is formed
between the donor D, an electronegative atom, which induces a
partially positively charge on its covalently bonded H, and acceptor
A, an electronegative atom with a lone pair, giving D
δ
–H
δ+
:
A
δ
, an electrostatically attractive situation. The H-bond donor
and acceptor in biological macromolecules is in most cases N or
O, but S–H and even C–H are seen to act as donors and make low
energy or weak hydrogen bonds (11).
Accounting for H-bonds in protein–ligand interactions is not as
simple as one might think, because they may encompass a wide
variety of chemical groups interacting via a hydrogen atom. In
fact, H-bonds form a continuum of interactions from those that
are barely distinguishable from dispersion alone, through to polar
and electrostatic charge interactions and on to almost complete
formation of a covalent bond. A number of classification schemes
have emerged (12). Perhaps the simplest is to divide them into three
classes according to the strength of the bond: weak (<4 kcal/mol),
moderate (4–15 kcal/mol), and strong (15–45 kcal/mol) accord-
ing to the nature of donor and acceptor groups (13). (N.B. strong
H-bonds are very rare in protein–ligand interactions). However,
Gilli et al. (12) emphasize that for any given D–H:A system the
energy may vary substantially for a given environment. For example,
in water the dimer HOHOH
2
is typically less than 5 kcal/mol,
but in acidic or basic environments, we have protonated
[H
2
OHOH
2
]+ or deprotonated [HOHOH]
forms, with energies reaching 32 kcal/mol. They conclude that
482 Paul J. Gane and A.W. Edith Chan
although the electronegativity of the donor and acceptor atoms
guides H-bond strength, it is controlled in many situations by
their acid–base dissociation constants.
Changes in hydrogen bonding number and strength are almost
ubiquitous in the formation of protein–ligand complexes. It would
seem, therefore, that H-bonds are likely to be important to ligand
binding; however, the energetic contribution they give to affinity
may be much less than is often assumed because H-bonds made
with a ligand often replace those previously made with water (14).
H-bonds are, however, extremely important for directionality and
recognition in protein–ligand interactions (15). Also, the dipolar
character of H-bonds may mean that H-bond formation preferen-
tially reduces off-rates of ligands. H-bonding ligands displaced 1 or
2by thermal fluctuations may be restrained by the 1/r
3
potential
and come back into contact rather than escaping into the bulk
solvent (Fig. 2).
1.3 Halogen Bonds The synthetic chemist has long added halogens to compounds with
the intent of improving ADME properties and they were also
thought to provide hydrophobic bulk. For some time it has been
observed that direct interactions (X-bonds) exist between organic
halides (X-bond donors) and protein electronegative atoms such as
O, S, or N (X-bond acceptors) forming, for example, a C–XO
bond. At first sight this is counterintuitive given the high electro-
negativity of the halogens, however, for Cl, Br, and I the electrons
in the C–X bond are asymmetrically distributed such that the
electrons surround the axis of the bond, but are withdrawn from
its terminus, producing a partial positive charge known as a sigma
hole at the halogen atom. This effect becomes larger with increas-
ing size of the halogen. The partial positive end that can interact
Fig. 2 Hydrogen bonding in protein–ligand interactions. (a) Weak and moderate strength H-bonds between
sildenafil (Viagra) and its protein target phosphodiesterase 5A (PDB entry 1TBF). In addition to H-bonds to
polar amino-acid side chains, weak bonds are identified to the aromatic heterocycle of sildenafil. (b) Many
moderate and strong H-bonds involving charged protein groups (lysine, arginine) mediate the interactions
between the cholesterol lowering drug fluvastatin and its target HMG-CoA reductase (PDB entry 1HWI)
Molecular Fields in Ligand Discovery 483
with a lone pair or aromatic π-cloud (showing parallels with
H-bonds) results in close approach of halide and electronegative
atoms (see Fig. 3a). Fluorine’s small size and extremely high elec-
tronegativity prevents it from forming halogen bonds. Parisini et al.
(16) highlight examples of X-ray crystal structures of small haloge-
nated molecules including drugs bound to proteins with these
directional halogen bonds. Calculating the electrostatic potential
surrounding the halogen containing compounds requires quantum
mechanical methods as shown by Riley and Hobza (17), who also
describe further examples from the PDB.
1.4 Aromatic
Interactions
Aromatic systems, such as benzene, are defined by delocalized π
electrons above and below the ring of bonded atoms, sandwiching
partial positively charged σbonds around the circumference of the
ring. This simple model is sufficient to explain the interactions
between aromatic rings (as a consequence of ππinteractions) and
of other functional groups interacting with aromatic πrings (18).
A benzene dimer is observed crystallographically and is largely
in two forms, slipped-parallel and T-shaped. Direct overlay of one
ring onto another is unstable, due to the electrostatic repulsion of
the πelectrons in the two molecules. The major component of
attraction appears to be dispersion forces with electrostatics
controlling the molecules’ orientation. These dispersion and elec-
trostatic terms probably control other aromatic interactions. A
good review is given by Tsuzuki (19) who describes not only ππ
interactions, but also OH–π, NH–π, CH–π, and cation–π
Fig. 3 Other electrostatic interactions in protein–ligand interactions. (a) A halogen bond between ethacrynic
acid and a tyrosine of human glutathione transferase P1-1 identified by the short distance (2.87 ) between
the tyrosine OH and chlorine substituent of the phenolic ring (PDB entry 3N9J). (b) Cation–πinteractions (small
spheres) between biopterin and nitric oxide synthase (PDB entry 1DMI). The positive guanidinium-group
of arginine and a positive nitrogen of the pteridine ring are attracted to delocalized πelectrons in aromatic
rings (c) Aromatic–aromatic interactions create offset stacking in the amodiaquine and histamine methyl-
transferase complex (PDB entry 2AOU)
484 Paul J. Gane and A.W. Edith Chan
interactions, which involve electrostatic attraction to the πelectrons
and are also of interest in protein–ligand and drug design applica-
tions (20). Functional substituents to the aromatic group can have
electron-withdrawing or electron-donating abilities, which modu-
late the πinteractions. An interesting, if unusual, example is hexa-
fluorobenzene which has an inverted electron distribution, due to
the highly electronegative fluorine atoms, will parallel stack with
benzene with a high energy of interaction of 8 kcal/mol. Aromatic
heterocycles, which are found in the majority of current small
molecule drugs, have distinct electronic properties that may require
special consideration (19).
With the exception of the cation–πinteraction, aromatic inter-
actions are energetically only one half to one quarter the value of a
weak H-bond, but they are nonetheless important in orientating
and directing of ligands within a binding site (see Fig. 3b, c)(21).
1.5 Hydrophobic
Interactions and the
Energetics of Binding
Water, the “lubricant of life” (22) is not merely a passive solvent in
which physicochemical interactions occur, but substantially contri-
butes to the balance of forces/energies which compete during
ligand binding. Water molecules which surround the ligand and
occupy the binding site are displaced to allow direct association
between the protein and the ligand. Changes in the interactions of
these displaced water molecules, consequently contribute to bind-
ing energetics. Water molecules which interact with solutes are
considered to be more ordered than bulk water. Hence, when
they are displaced they are thought to provide a large positive
entropic component which favors binding. In addition, the nonpo-
lar surfaces of ligand and protein interact via dispersion forces
which although individually weak are additive and sum over sub-
stantial surface areas providing a large enthalpic component to
binding. The ligand, when bound, is in a fixed conformation
along with the interacting residues of the protein—this loss of
mobility lowers the entropy and directly opposes binding. The
more tightly bound the ligand, the greater the enthalpy of binding
and the more restricted its motion and the higher the entropic
penalty. This trade-off between the two components of the free
energy of binding is known as entropy–enthalpy compensation.
Detailed features of the interactions in a particular protein–li-
gand complex are significant in determining the net contribution of
water to binding. Other factors to consider are the size of the ligand
and its shape in relation to the geometry of the binding cavity and
the surface it presents to interfacial water (water molecules trapped
between ligand and protein) (23). Myslinski et al. (24) conclude
that displacing bound water does not necessarily result in favorable
entropy to binding if particular bound water molecules are less
ordered than is generally thought.
It is also important to remember that in solution both water
and ions dramatically reduce any long-range Coulombic
Molecular Fields in Ligand Discovery 485
interactions, and that conversely the environment found within
certain binding sites may enhance the electrostatics, altering pK
a
s
and chemical reactivity.
1.6 Conclusion Although, we have more or less sophisticated ideas of the types of
interactions that are significant in the association of proteins and
small molecules, the prediction of the free energy of binding of
small-molecule ligands to a protein is still an unsolved problem,
primarily because of the non-additivity and multiplicity of the
interactions involved (25). Particular difficulties are found with
calculating solvation–desolvation of both ligand and protein and
the role of counterions; flexibility and conformational changes;
long-range electrostatics; changes in ionization/pK
a
of ligand
functional groups and binding site residues and the kinetics asso-
ciated with some H-bonds (26). Given these difficulties and limita-
tions, it is remarkable how successful many of the computational
tools considered in the following sections are in aiding ligand and
drug discovery.
2 In Silico Methods for Ligand and Early-Stage Drug Discovery
The route from first identifying a protein target, to a useful
biological reagent may be long, and that to a marketed drug is
certainly long and tortuous with many feedback cycles. Since in
both ligand and drug discovery the initial goal is to identify a small
molecule with high affinity and specificity for a protein binding-site
many of the early stages of the process (Fig. 4) are similar.
2.1 Finding Hits
with Virtual Screening
High throughput screening (HTS), the testing of tens of thousands
to millions of compounds for activity against a chosen target, is the
most common strategy in the pharmaceutical industry for the initial
identification of small molecules that are active against a protein
target. However, HTS is immensely expensive in terms of time,
labor, and resources and is not possible in an academic environment
or small companies, and not justifiable outside of drug development
programs. To make matters worse the success rate of HTS is often
poor, so why make the haystack larger when looking for a needle?
A more “rational” approach is to assay a small subset of com-
pounds which have been preselected to have a much higher likeli-
hood of binding to the chosen target. This is most often achieved
with computational filtering or “virtual screening” of a large data-
base of available chemical compounds. There are many virtual
screening methods and many choices of program for each given
method.
Docking, fragment building and de novo ligand generation are
routine methods for using structure to screening large compound
libraries or to build novel ligands into binding sites (27).
486 Paul J. Gane and A.W. Edith Chan
TARGET
Identification,
Characterization,
Site analysis,
Mutations/SNPs
LIGANDS
Known natural substrate(s),
Ligand activity data
STRUCTURE
3D data, X-ray, NMR
SMALL MOLECULES
Commercial molecule suppliers,
Academic specialty collections,
Clinical/drug collections,
Fragment sets,
Target specific libraries
MOLECULAR DATABASES
Store small molecule data in
flat file or relational DB
FILTERS
Filter database by chemical descriptors, reactive
groups, frequent hitters, remove salts/metals etc.
STRUCTURE BASED DRUG DESIGN
Docking, de novo design, Fragment
Based Drug Design, Grid methods, etc.
LIGAND BASED DRUG DESIGN
Substructure/similarity searching,
(Q)SAR, Pharmacophore, CoMFA, etc.
SCREENING
Biological / biochemical / pharmacological / functional assays
HITS
Initial hits - reconfirm
and obtain Dose
Response Curves
EXPAND HITLIST / IDENTIFY KEY SCAFFOLDS
Search commercial sources for analogues around hits,
Bioisostere substitutions,
Molecular Field Comparisons
PURCHASE ANALOGUES
Chemical suppliers
CHEMICAL SYNTHESIS
Decorate scaffolds and generate analogues
RESCREEN
Rescreen expanded hitlist
ANALYSIS
Determine (Q)SAR,
ADMET prediction
HIT TO LEAD
Select promising leads
ACTIVITY STUDIES
Cell / Tissue / Animal studies, ADMET
CLINICAL TRIALS
PI, PII, PIII, PIV
Fig. 4 Flow diagram of the drug discovery process. A similar process applies to the discovery of any potent
ligand but note that many feedback cycles occur throughout the process and the later stages greatly simplified
Molecular Fields in Ligand Discovery 487
Their great advantage is in discovering new scaffolds, which may
overcome possible patent and synthetic issues, but they do require
an X-ray, NMR, or homology model of the 3D structure of the
target protein (see Chapter 19 for more detail on these methods).
In contrast, traditional “pharmacophore” searching is “ligand-
based” where a number of known active compounds are super-
posed by optimizing the overlap of simple descriptors of their
chemical groups and/or interaction-capable sites. Regions of struc-
ture or functionality which are in common are regarded to be
important for activity and a pharmacophore point having particular
properties is placed there. Because many molecules are rather flexi-
ble, successfully matching molecules from a database to a set of
pharmacophore points requires that sufficient conformations of
each molecule have been generated to adequately sample the possi-
ble spatial distributions of functionality.
A number of the top scoring compounds from a virtual screen
are chosen for a preliminary biochemical/biological assay. Active
compounds or “hits” are subject to further assays to confirm activity.
2.2 Expanding and
Diversifying Hits
It is usually the case that the number of experimentally confirmed
HTS hits is small or the core molecular unit (better known as the
scaffold) of each of the hits or known binders is chemically very
similar. Increasing the number of hits can sometimes simply be a
matter of purchasing analogues which share the scaffold and re-
assaying. An online tool, ZINC (28) has revolutionized the pur-
chase of compounds by collating over 100 different vendors on a
single website with an integrated similarity and substructure search
applet.
The results of variation on a scaffold can point to fragments of
the molecules which are important for activity and those which are
not, making it possible to build a structure–activity relationship
(SAR) for that series.
A drug design project, in particular, is more likely to succeed if
there is more than one scaffold library, since diversity gives a greater
probability of later avoiding problems with toxicity and metabo-
lism. Identifying different structural scaffolds with similar
biological activities is known as scaffold hopping. Pharmacophore
searching is often used for scaffold hopping as it compares generic
properties such as location of H-bond donors/acceptors, presence
of aromatic rings, full/partial charges, hydrophobicity, etc., rather
than their specific chemical constituents. Li et al. (29) propose a
measure of scaffold hopping by quantifying the differences between
scaffolds which may be of great benefit in deciding on the diversity
of scaffolds to select.
2.3 Bioisosteres Bioisosteric replacement is one scaffold hopping method. A bioi-
sostere is a molecular fragment capable of replacing another whilst
retaining similar biological function of the molecule. Bioisosteres
488 Paul J. Gane and A.W. Edith Chan
may also be known as nonclassical isosteres to differentiate them
from the stricter definition of an isostere which requires a chemical
and a physical match such as volume and shape. There are many
known “standard” bioisosteres for all common functional groups as
well as ring equivalents, peptide bond substitutions, isomer and
bond inversions; for an excellent and thorough guide see Wermuth
(30) (pp 290–342), also summary tables of bioisosteres are given in
David Young’s book (5) (pp 60–64). It is important to recognize
that some bioisosteres will work well with some protein targets but
not others. Molecular field software has proven useful in discover-
ing new bioisosteres and is exemplified in the program FieldStere
(see Subheading 3.3).
2.4 Molecular Fields
in Virtual Screening
The complementary matching of electrostatic, steric, and hydro-
phobic properties of a ligand to its protein binding site is the basis
of drug action. Drugs with similar structures usually have similar
biological functions; this is the basis of drug design. It is well
known that some compounds with different chemical structures
can have the same biological effect; this is the basis of bioisosteric
replacement; and is possible despite chemical differences because
molecules’ electrostatic and hydrophobic properties may in fact be
similar. Comparing different compounds by the spatial distribution
of their properties, i.e., their “molecular fields” rather than by
structure alone, has resulted in the discovery of compounds having
different core structures (scaffolds) but comparable biological
activity (31).
The challenge is to calculate the molecular field properties
quickly and to score them quantitatively so that they accurately
distinguish differences and similarities between one molecule and
another. However, fields vary continuously with conformation and
across the molecular surface making comparisons between mole-
cules extremely difficult. Fields can be sampled at defined points on
a 3-dimensional grid and compare values at grid points. However,
the grid has to be at a high enough resolution to capture rapid
changes in the field (this is very apparent in protein electrostatics
calculations (32)) and consequently large amounts of grid data are
generated and computational comparison is expensive.
A solution to the comparison problem is to look only at the
extrema (minima and maxima) of these fields where intermolecular
interactions are most likely to occur (33). The IsoStar database
catalogues the likely interaction of chemical functional groups
between small molecules and between ligands and proteins (34).
The calculation and comparison of extrema is embodied in the
Cresset (35) collection of programs based on the extended electron
density (XED) force field by Vinter (36). The majority of molecular
mechanics force fields mimic the unequal distribution of electrons
in a molecule by applying atom-centered partial charges. XED
constructs a negative set of electron pseudo-orbitals around the
Molecular Fields in Ligand Discovery 489
positively charged nuclei to give neutral atoms and then distributes
partial charges appropriately at the orbital locations to mimic the
electron distribution. This gives a much more accurate description
of the molecular electrostatic potential, particularly of π-clouds and
lone pairs.
The Cresset definition of extrema is called Field Points, and the
number calculated per molecule is approximately equal to the heavy
atom count. This is a vastreduction in data points compared to using
the full electrostatic field grid, making this a computationally tracta-
ble approach to in silico HTS. The paper by Cheeseright et al. (37)
details the molecular mechanics potentials used in calculating the
steric (van der Waals), Coulombic, and hydrophobic fields via molec-
ular “probes” first demonstrated by the ground-breaking work of
Goodford (38) in visualizing protein–ligand interactions (Fig. 5).
Virtual screening requires that once the Field Points have been
calculated for a given molecule they need to be compared and
scored to each member of the chemical library which is to be
Fig. 5 The diuretic amiloride represented (a) as a stick model, (b) electrostatic
contours (dark/red positive, light/cyan negative), (c) reducing the fields to Field
Points, (d) final display of Field Points including smaller spheres representing van
der Waals attractive and hydrophobic interaction sites. Sphere size is in propor-
tion to the magnitude of the extrema
490 Paul J. Gane and A.W. Edith Chan
screened. The set of Field Points are compared between two
molecules by sampling conformations to maximize the overlay of
the fields and find the optimal alignment. This is in contrast to
aligning molecules by structure; here it is the fields that are com-
pared, and therefore, it should find matches between molecules of
very different structural classes but having a similar number and
distribution of Field Points. Field Point techniques are thus per-
fectly suited for scaffold hopping.
Cresset has developed a commercial package of programs with
provision for free teaching licenses and substantial academic dis-
counts. This approach by software vendors, others of note are
OpenEye (39), MOE (40), CCDC (41), etc., allows even modestly
funded academic labs or institutions to establish ligand and early
stage drug discovery programs (42).
3 Methods
This section will focus on the Cresset suite of software aimed at
discovering small molecules by virtual screening, scaffold hopping
and identifying new bioisosteres. It is designed to be an introduc-
tion for chemists and modellers unfamiliar with the software, not an
exhaustive description of every parameter, which are detailed in
each of the Cresset Software Manuals.
Three of the Field programs from Cresset (FieldTemplater,
FieldAlign, and FieldStere) are standalone and can be run on
most platforms. FieldScreen, however, requires a Linux cluster
and extra software tools, it will not be discussed further here, but
Cresset do provide a number of unique options for running
FieldScreen which as the name suggests, is a high throughput
screening tool. A separate visualization program, FieldView is freely
available for calculating and viewing fields and Field Points.
FieldTemplater is an advanced pharmacophore matching tool
coupled with conformer generation. If the bioactive conformation
of a ligand (or ligands) is known it can be used as a 3D template to
compare with other molecules. FieldTemplater assumes that all
ligands bind in a similar fashion to the template. In many projects
there is no known conformation to work from, but aligning the
Fields of a few active compounds reveals Field Points in common
and suggests those regions which are responsible for activity. Thus,
FieldTemplater performs conformational searching to optimize the
overlap of Field Points and predict the bioactive conformations.
Any matches can be confirmed by visualizing their Field Points.
These conformations can then be used in the FieldAlign program.
FieldAlign will align a small library of compounds (up to 500)
against a single conformation of a known 3D structure or against a
prealigned set from FieldTemplater, and rank them according to
their Field Point similarities. Because of the computational expense
Molecular Fields in Ligand Discovery 491
of 3D field alignment, for large compound libraries of many
thousands to millions of molecules other screening methods need
to be used first such as docking or traditional pharmacophore
filtering. These preselected hits can be then rescreened with Fiel-
dAlign to obtain an alternative and potentially more accurate rank-
ing using molecular field similarity.
FieldStere is a bioisostere search routine, which goes beyond
the traditional functional replacements substituting selected por-
tions of your molecule with its own database of Field fragments and
comparing the total fields of the new molecule and the parent, with
the aim of retaining the integrity of the original activity. FieldStere
is especially useful in proposing diverse substitutions which are not
apparent unless the Field Points are superposed.
3.1 FieldTemplater Input: A file of known ligands in mol, mol2, or sdf format
(see Note 1).
lStart FieldTemplater (a Wizard window should appear, use this
for the first trial).
lYou are asked to load one or more molecules via the Import bar.
lOnce these have been added, click Next. Constraints can be
added to each molecule to enforce overlay of particular func-
tional groups in the situation where you know those groups
interact with the same region of the binding site.
lNow you can select the “quality” level of the r un. This refers to the
extent of shortcuts taken in the calculation protocols, with higher
quality settings corresponding to greater sampling of conforma-
tions and more precise field alignments. Each of the settings
(quick, normal, accurate) are appropriate to different scenarios,
e.g., quick may be adequate for a fragment or to get an initial filter
of a complex system prior to running the more computationally
demanding and time consuming “accurate” setting.
lThe results appear as sets of conformations ranked according to
field similarity, shape similarity or combined similarity. Field
similarity is based purely on the electrostatic match of two
molecules, whereas shape similarity is based on their Gaussian
volume overlap. Both similarity measures have their merits, but
as the contribution to binding of electrostatics and shape will be
dependent of the chemotype of the ligand and the nature of the
target often the combined score is most useful.
lYou can save the project and then visually inspect the superpo-
sition, manually making different templates visible or not. It is
common to manually select templates that have high similarity
scores or match some other data on known binding mode or
simply through chemical intuition. Finally “visible” templates
can be selected for export to an sdf file. If “visible” is not
selected it will save all generated conformers (Fig. 6).
492 Paul J. Gane and A.W. Edith Chan
3.2 FieldAlign lInput a single or small set of reference structures preferably in
its/their bioactive conformation(s) or a computed conforma-
tion from FieldTemplater as sdf, mol, mol2, or SMILES format
(see Note 2).
lNext add a database in sdf (see Note 3 on database preparation).
For the example in Fig. 7a filtered version of the John Hopkins
Clinical Collection (43) of known drugs was used.
lProtonation and tautomerization are issues of which the user
must be aware. FieldAlign has built in routines for automatic
Fig. 6 FieldTemplate creation with Cresset software. The interaction FieldPoints for four α2 adrenergic
antagonists (left) can be used to superimpose them (right) despite their belonging to two distinct chemical
classes
Fig. 7 Field Align. Searching a database of existing drugs with the template created from the four molecules in
Fig. 18.6 identifies bendazol (an antihelmintic) as the top hit (bottom center). Bendazol is superposed with the
four reference molecules (bottom right)
Molecular Fields in Ligand Discovery 493
protonation, but where experimental information is available it
is better to specify the protonation state manually using the
software’s molecular editor window. Tautomers are not dealt
with automatically, but can also be manually specified in the
molecular editor—this is best done by including any possible/
likely tautomers as separate molecular entities. The screening
outcome may be very sensitive to tautomers as they invariably
produce different field patterns.
lFieldAlign ranks the members of the input database according
to their Field similarity score with the reference molecule(s) or
template.
lClicking the + symbol expands different conformational hits for
that molecule.
lHighlighting any molecule from the database given in the left-
hand window immediately places it into the main viewing win-
dow aligned with the reference(s). This enables straightforward
visual inspection.
FieldAlign is not a “high throughput” screening program, and will
only process 500 molecules per run. It is best to use FieldAlign
on pre-filtered sets of molecules, e.g., as a post-docking/
pharmacophore screen. FieldAlign will provide matches with dif-
ferent chemical functionality from that of the reference(s) some of
these new scaffolds will potentially have similar biological activity.
3.3 FieldStere FieldStere is a fragment replacement program which compares
the Fields of its internal list of fragments with a selected part of
the input molecule of interest. If a field match is found, it replaces
the selected part with the new fragment and field scores and com-
pares the entire field of the new molecule with the entire field of the
original input molecule.
lInput a reference molecule (again in sdf, mol, mol2, or SMILES
format).
lYou can allow FieldStere to choose the protonation state.
lDrag the cursor around the section of the molecule you wish to
replace.
lNow select the database(s) to include in your search, there are five
to choose from very commonly found fragments to very rare
ones. The more databases that are included the longer the proces-
sing takes. There are a number of scoring and filtering parameters
that can be adjusted by the user before running FieldStere, e.g.,
which atoms to or not to use to reconnect fragments, the size of
fragments to consider, their logP, number of rotatable bonds,
whether fragments should be incorporated into rings, etc.
The results are presented in multiple windows (Fig. 8), the
main graphics display the original reference molecule overlaid or
494 Paul J. Gane and A.W. Edith Chan
Fig. 8 FieldStere—bioisostere replacement based on fragment selection. (a) Drag cursor around chosen
fragment, an imidazoline which will be replaced by its steric and electrostatic equivalents (as determined by
FieldPoints) drawn from the built-in chemical group database. Bioisostere selection is highly customizable.
(b) Settings can be adjusted in a series of menus for scoring, size, filters, etc. Filters are useful for selecting
particular chemotypes and are comparable to pharmacophore constraints. (c) The reference molecule
(left) with a modified version in which the fragment is replaced by a selected bioisostere (right). (d) Thumbnails
of the all matching fragments are displayed in a sorted list along with physicochemical data
Molecular Fields in Ligand Discovery 495
side by side with its new fragment. The candidate fragments are
ranked based upon the similarity score, when selected they appear
in the graphics window. Automation of the above processes is
available (see Note 4). Selection of these modified molecules is
again usually manual, although their similarity scores to the original
molecule are given and could be used as a filter in an automated
procedure. Other factors significant in selection may include the
torsional strain energy created by the fragment substitution (which
relates to the likelihood of conformation being adopted by the
substituted molecule), chemical tractability and novelty.
As an idea generator FieldStere is an excellent tool and often
output can be checked against known SAR for validation.
3.4 Conclusion The Field software can be applied to scaffold hopping, expanding
a set of known hits/leads by bioisostere replacement, post-
processing of virtual screening results, bioactive conformer genera-
tion, and many other applications which require new ideas to
suggest new chemistry or to overcome existing patents.
Of course, given the many software products available, some
will be superior for specialist functions, but it is the focus on the
importance of molecular field extrema which allows these particular
programs to rapidly generate unique solutions to molecular
similarity.
4 Notes
1. Mol and mol2 formats describe atomic composition, bond
connectivity, and coordinates (44), originally developed by
Molecular Design Ltd and Tripos, they are now an industry
standard. Structure data format (sdf) files are simply concate-
nated mol or mol2 format entries separated by a line containing
$$$$. If you are downloading a ligand from a PDB/RCSB
entry scroll down to the Ligand Chemical Component box
and click the Download pulldown and save as an sdf. Note
this will include x,y,z, coordinates as well as SMILES and
InChI string representations of the molecular architecture.
Alternatively you can extract a ligand from a pdb format by
using a text editor or from modelling software and modify the
file format.
2. Where a receptor/protein structure is known “excluded
volumes” can also be used as a steric boundary during FieldA-
lign or FieldStere runs. When available such boundaries can be
used to prevent selection of molecules or fragments which
would otherwise clash with protein atoms.
3. Molecular databases must be prepared beforehand using filters
to limit molecular weight. This is particularly important if your
496 Paul J. Gane and A.W. Edith Chan
database of structures came from a general chemicals catalogue
or a set of therapeutics which cover large peptides and even
proteins. It is best to set an upper and lower MW filter as this
will also remove salts, metals and other counterions which are
not usually of interest. Chemically reactive compounds and
those which are known to interfere with assays are usually
removed as are a group molecules known as “frequent hitters,”
which are known to bind to many protein families and may
diminish selectivity, “privileged scaffolds,” commonly occur-
ring core structures may be omitted or retained depending on
synthetic or patent strategies (45). Which compounds are
included or excluded may vary according to established proce-
dure, personal chemical bias, or the protein target under con-
sideration.
4. Non-graphical command-line tools are also available for Field-
Stere, FieldAlign, and the XedTools (see Note 5). These
command-line tools provide added control, increased speed,
and automation of repeated applications of procedures to dif-
ferent ligands. Field and Xed programs are also integrated into
data analysis pipeline tools such as Knime (http://www.knime.
org) and PipelinePilot (Accelrys, Inc.).
5. XedTools provide forcefield based manipulation of molecules.
Xedmin uses the XED force field to energy minimize ligand
conformations with or without the protein present. XedeX
generates diverse families of distinct, energy minimized con-
formations of small molecules for use in subsequent Field
template construction or alignment.
Acknowledgments
OpenEye for free access to their full suite of programs, MOE and
GOLD for their generous academic pricing, Cresset for their free
teaching licenses and support. Andy Vinter and Martin Slater for
their excellent advice.
References
1. Nicholls A (2011) What do we know? Simple
statistical techniques that help. In: Bajorath J
(ed) Chemoinformatics and computational
chemical biology. Springer Science, New York
2. Merz KM et al (2010) Drug design structure-
and ligand-based approaches. CUP, Cam-
bridge
3. Leach AR (2001) Molecular modelling: princi-
ples and applications, 2nd edn. Longman,
Edinburgh
4. Gasteiger J, Engel T (2004) Chemoinfor-
matics: a textbook. Wiley, Wienheim
5. Young DC (2009) Computational drug
design: a guide for computational and medici-
nal chemists. Wiley, New Jersey
6. Rigby M et al (1986) The forces between
molecules. OUP, Oxford
7. Stone AJ (2008) Intermolecular potentials.
Science 321:787–789
Molecular Fields in Ligand Discovery 497
8. Bissantz C, Kuhn B, Stahl M (2010) A medici-
nal chemist’s guide to molecular interactions. J
Med Chem 53:5061–5084
9. Chan AW, Laskowski RA, Selwood DL (2010)
Chemical fragments that hydrogen bond to
Asp, Glu, Arg and His sidechains in protein
binding sites. J Med Chem 53:3086–3094
10. Protein DataBank (PDB) at RCSB: http://
www.rcsb.org/
11. Williams MA, Ladbury JE (2003) Hydrogen
bonds in protein–ligand complexes. In: Bo
¨hm
HJ, Schneider G (eds) Protein–ligand interac-
tions from molecular recognition to drug
design. Wiley, Weinheim
12. Gilli G, Gilli P (2009) The nature of the hydro-
gen bond: outline of a comprehensive hydro-
gen bond theory. OUP, Oxford
13. Labowski SJ (2006) Hydrogen bonding: new
insights. Springer, Netherlands
14. Foloppe N (2005) Structure-based design of
novel Chk1 inhibitors: Insights into hydrogen
bonding and protein–ligand affinity. J Med
Chem 48:4332–4345
15. Kubinyi H (2008) The changing landscape in
drug discovery. In: Stroud RM, Finer-Moore J
(eds) Computational and structural approaches
to drug discovery. RSC, Cambridge
16. Parisini E et al (2011) Halogen bonding in
halocarbon–protein complexes: a structural
survey. Chem Soc Rev 40:2267–2278
17. Riley KE, Hobza P (2011) Strength and char-
acter of halogen bonds in protein–ligand com-
plexes. Cryst Growth Des 11:4272–4278
18. Hunter CA, Sanders JKM (1990) The nature
of ππInteractions. J Am Chem Soc
112:5525–5534
19. Tsuzuki S (2005) Interactions with aromatic
rings. Struc Bond 115:149–193
20. Gooding M (2011) Exploring the interaction
between siRNA and the SMoC biomolecule
transporters: implications for small molecule-
mediated delivery of siRNA. Chem Biol Drug
Des 79:9–21
21. Lanzarotti E (2011) Aromatic–aromatic inter-
actions in proteins: beyond the dimer. J Chem
Inf Model 51:1623–1633
22. Barron LD, Hecht L, Wilson G (1997) The
lubricant of life: a proposal that solvent water
promotes extremely fast conformational fluc-
tuations in mobile heteropolypeptide struc-
ture. Biochemistry 36:13143–13147
23. Setny P, Baron R, McCammon JA (2010) How
can hydrophobic association be enthalpy
driven? J Chem Theory Comput 6:2866–2871
24. Myslinski JM et al (2011) Protein–ligand inter-
actions: thermodynamic effects associated with
increasing nonpolar surface area. J Am Chem
Soc 133:18518–18521
25. Hassan SA et al (2005) Computer simulation
of protein–ligand interactions: challenges and
applications. In: Nienhaus GU (ed) Protein–li-
gand Interactions. Humana, New Jersey
26. Schmidtke P et al (2011) Shielded hydrogen
bonds as structural determinants of binding
kinetics: application in drug design. J Am
Chem Soc 133:18903–18910
27. Cavasotto CN, Phatak SS (2011) Docking
methods for structure-based library design.
In: Zhou JZ (ed) Chemical library design.
Springer, New York
28. Irwin JJ, Shoichet BK (2005) ZINC a free
database of commercially available compounds
for virtual screening. J Chem Inf Model
45:177–182
29. Li R et al (2011) Development of a method to
consistently quantify the structural distance
between scaffolds and to assess scaffold hopping
potential. J Chem Inf Model 57:2507–2514
30. Wermuth CG (2008) The practice of medicinal
chemistry. Academic, London
31. Low CMR (2005) Scaffold hopping with
molecular field points: identification of a
cholecystokinin-2 (CCK2) receptor pharmaco-
phore and its use in the design of a prototypical
series of pyrrole- and imidazole-based CCK2
antagonists. J Med Chem 48:6790–6802
32. Warwicker J (1994) Improved continuum elec-
trostatic modelling in proteins, with compari-
son to experiment. J Mol Biol 236:887–903
33. Apaya RP (1995) The matching of electrostatic
extrema: a useful method in drug design? A
study of phosphodiesterase III inhibitors. J
Comput Aided Mol Des 9:33–43
34. Bruno IJ et al (1997) Isostar: a library of infor-
mation about non-bonded interactions. J
Comput Aided Mol Des 11:525–537, http://
www.ccdc.cam.ac.uk/products/csd_system/
isostar/
35. Cresset software http://www.cresset-group.
com
36. Vinter JG (1994) Extended electron distribu-
tions applied to molecular mechanics of inter-
molecular interactions. J Comput Aided Mol
Des 8:653–668
37. Cheeseright T et al (2006) Molecular field
extrema as descriptors of biological activity:
definition and validation. J Chem Inf Model
46:665–676
38. Goodford PJ (1985) A computational proce-
dure for determining energetically favorable
binding sites on biologically important macro-
molecules. J Med Chem 28:849–857
498 Paul J. Gane and A.W. Edith Chan
39. OpenEye http://www.eyesopen.com/
40. MOE (Molecular Operating Environment)
http://www.chemcomp.com/software.htm
41. GOLD http://www.ccdc.cam.ac.uk/pro-
ducts/life_sciences/gold/
42. Gane PJ, Chan E, Selwood D. ChemiBank
http://www.ucl.ac.uk/chemibank
43. John Hopkins clinical collection http://htc.
wustl.edu/library/JHCCL.html
44. Mol2 format http://tripos.com/tripos_re-
sources/fileroot/pdfs/mol2_format.pdf
45. Rishton GM (2003) Non-leadlikeness and
leadlikeness in biochemical screening. DDT
8:86–96
Molecular Fields in Ligand Discovery 499
... In practise, the use of this approach allows the nuclei and electrons to be quantified into a sole atom-like particle. 75 83 Force fields are mathematical concepts that combine first-principles physics and parameter fitting to quantum mechanical calculations and empirical data to define molecules, represented as atoms connected by bonds with lengths, angles and energies. 84 There are different types of force fields developed with different levels of complexity designed to be used for different systems, schematically a force field being: ...
... The angle bending is treated the same manner as bond stretching using Hook's law through the deviation of the angles from a reference angle via the following equation: [83][84][85] ...
... It is estimated that a smaller energy is required to move the angle from the reference or equilibrium than to stretch a bond. [83][84][85] In bond stretching and angle bending the creation of any change or deformation from the equilibrium requires a large energy input, so that the main difference in structures and energies are due to the torsional and non-bonded interactions. The structural elements as well as the molecular geometry could be easily understood by defining the existence of barriers to rotation about bonds. ...
Conference Paper
Gaucher’s Disease (GD) is a rare recessive disorder produced by the dysfunction of the lysosomal enzyme Glucocerebrosidase (GCase). GCase catalyses the cleavage of the glycolipid Glucosylceramide. The lack of functional GCase leads to the accumulation of its lipid substrate in lysosomes causing GD. GD presents a great phenotypic variation, symptoms ranging from asymptomatic adults to early childhood death due to neurological damage. More than 250 mutations in the protein GCase have been discovered that result in GD. Being able to link structural modifications of each mutation to the phenotypic variation of GD would enhance the understanding of the disease. The aim of this work is to understand the structural dynamics of wild type and mutant GCase. A model of the complex of the enzyme GCase with its facilitator protein, Saposin-C (Sap-C) was generated using Protein-Protein docking (PPD). In this work, a knowledge-based docking protocol that considers experimental data of protein- protein binding has been carried out. Here, a reliable model of the enzyme GCase with its facilitator protein is presented and is consistent with the experimental data. To understand the structural mechanism of function of the enzyme GCase, it was imperative to study its structural dynamics and conformational changes influenced by its interaction with other components including lipid bilayer, facilitator protein or substrate. Coarse-Grained MD (CG-MD) was employed to study lipid self-assembly and membrane insertion of the complex. Classical Atomistic MD (AT-MD) was used to study the dynamics of the interactions between different components of the simulation. Furthermore, the results of ten different AT-MD simulations sampling 9 s have been analysed. An activation method of GCase by Sap-C has been proposed, the change in conformation of GCase when its facilitator protein is present has been highlighted, through the stabilization of the loops at the entrance of the binding site. The differences in protein-protein binding when GCase is mutated have also been emphasised. Finally, Anharmonic Conformational Analysis and Markov State Models have been used to build a kinetic model of the system. This model supports our activation mechanism hyphothesis.
... Fig. 1 displays the field points template features of SARS-CoV-2 repurposed drugs by using Forge visualization tool. As these field points generate the four diversified molecular fields containing positive and negative electrostatic potential surface, shape and hydrophobic field surface [49][50][51]. ...
... For inhibiting the COVID-19 main protease, pharmacophore field templates were generated for the FDA approved SARS-CoV-2 repurposed drugs comprising Nafamostat, Hydroxyprogesterone caporate and Camostat mesylate. These drugs after field pharmacophore generation were now bioactive conformations which help in the identification of common pharmacophores features of marine drugs [51]. Fig. 2 showed the negative field points in cyan color which generated due to the interactions of positive regions or H-bond donors of receptor target with molecule. ...
Article
Full-text available
Since the inception of COVID-19 pandemic in December 2019, socio-economic crisis begins to rise globally and SARS-CoV-2 was responsible for this outbreak. With this outbreak, currently, world is in need of effective and safe eradication of COVID-19. Hence, in this study anti-SAR-Co-2 potential of FDA approved marine drugs (Biological macromolecules) data set is explored computationally using machine learning algorithm of Flare by Cresset Group, Field template, 3D-QSAR and activity Atlas model was generated against FDA approved M-pro SARS-CoV-2 repurposed drugs including Nafamostat, Hydroxyprogesterone caporate, and Camostat mesylate. Data sets were categorized into active and inactive molecules on the basis of their structural and biological resemblance with repurposed COVID-19 drugs. Then these active compounds were docked against the five different M-pro proteins co-crystal structures. Highest LF VS score of Holichondrin B against all main protease co-crystal structures ranked it as lead drug. Finally, this new technique of drug repurposing remained efficient to explore the anti-SARS-CoV-2 potential of FDA approved marine drugs.
... Gane et al. demonstrated that the biological effects of compounds are associated with their chemical structure. Although the compounds with similar biological effects usually have identical structures, some compounds with dissimilar structures show comparable biological outcomes too (71). Interestingly, we realized that ZINC73408075 and ZINC06482373 compounds have identical structure and their interactions with the residues in binding site2 of Wnt2 were alike too ( Figures S1F and S1H, Supplementary File). ...
Article
Full-text available
Wnts are the major ligands responsible for activating Wnt signaling pathway through binding to Frizzled proteins (Fzd) as the receptors. Among these ligands, Wnt2 plays the main role in the tumorigenesis of several human cancers especially colorectal cancer (CRC). Therefore, it can be considered as a potential drug target. The aim of this study was to identify potential drug candidates against two binding sites of Wnt2. Structure-based virtual screening approaches were applied to identify compounds against binding sites of Wnt2 for inhibiting the interaction Wnt2 and Fzd receptors. The best hit compounds from molecular docking of National Cancer Institute diversity set II database were used for structural similarity search on ZINC database, obtaining large hit compounds query to perform a virtual screening and retrieving potential lead compounds. Eight lead compounds were selected while their binding affinity, binding modes interactions, and molecular dynamics simulations studies were assessed. Molecular docking studies showed that eight selected lead compounds can bind to the desired binding sites of Wnt2 in a high affinity manner. Bioavailability analysis of the selected lead compounds indicated that they possessed significant drug like properties. Thus, these lead compounds were considered as potential drug candidates for inhibiting Wnt signaling pathway through combining with the binding sites of Wnt2 and hindering the interaction of Wnt2 and Fzd receptors. Our findings suggest that Wnt2 binding sites may be a useful target for treatment for CRC fueling the future efforts for developing new compounds against Wnt signaling pathway.
Article
Virtual screening has played a significant role in the discovery of small molecule inhibitors of therapeutic targets in last two decades. Various ligand and structure-based virtual screening approaches are employed to identify small molecule ligands for proteins of interest. These approaches are often combined in either hierarchical or parallel manner to take advantage of the strength and avoid the limitations associated with individual methods. Hierarchical combination of ligand and structure-based virtual screening approaches has received noteworthy success in numerous drug discovery campaigns. In hierarchical virtual screening, several filters using ligand and structure-based approaches are sequentially applied to reduce a large screening library to a number small enough for experimental testing. In this review, we focus on different hierarchical virtual screening strategies and their application in the discovery of small molecule modulators of important drug targets. Several virtual screening studies are discussed to demonstrate the successful application of hierarchical virtual screening in small molecule drug discovery.
Book
Now in its third edition, this classic reference is the one-stop-shop for information on the foundations of medicinal chemistry for pharmaceutical researchers who are involved in drug development & discovery but who do not have a background in medicinal chemistry. Wermuth aids pharmaceutical researchers and chemists in making faster, more accurate identifications of the active substances that could potentially treat the disorder they are researching. New chapters on Drug Absorption & Transport give pharmaceutical scientists information on how potential drugs can move through the drug discovery/development phases more quickly. This third edition still stands as the only source for practical aspects of medicinal chemistry by focusing on the daily problems met by the medicinal chemist in drug discovery. NEW TO THIS EDITION: * Focus on chemoinformatics and drug discovery * Enhanced pedagogical features * New chapters including: - Drug absorption and transport - Multi-target drugs * Updates on hot new areas: NEW! Drug discovery and the latest techniques NEW! How potential drugs can move through the drug discovery/ development phases more quickly NEW! Chemoinformatics.
Article
Hydrogen bond (H-bond) effects are well known: it makes sea water liquid, joins cellulose microfibrils in sequoia trees, shapes DNA into chromosomes, and polypeptide chains into wool, hair, muscles, or enzymes. However, its very nature is much less known and we may still wonder why O-H O energies range from less than 1 to more than 30 kcal/mol without evident reason. This H-bond puzzle is tackled here by a new approach aimed to obtain full rationalization and comprehensive interpretation of the H-bond in terms of classical chemical-bond theories starting from the very root of the problem, an extended compilation of H-bond energies and geometries derived from modern thermodynamic and structural databases. From this analysis new concepts emerge: new classes of systematically strong H-bonds (CAHBs and RAHBs: charge- and resonance-assisted H-bonds); full H-bond classification in six classes (the chemical leitmotifs); assessment of the covalent nature of all strong H-bonds. This finally leads to three distinct though inter-consistent theoretical models able to rationalize the H-bond and to predict its strength which are based on the classical VB theory (electrostatic-covalent H-bond model, ECHBM), the matching of donor-acceptor acid-base parameters (PA/pKa equalization principle), and the shape of the H-bond proton-transfer pathway (transition-state H-bond theory, TSHBT). A number of important chemical and biochemical systems where strong H-bonds play an important functional role are surveyed, such as enzymatic catalysis, ion-transport through cell membranes, crystal packing, prototropic tautomerism, and molecular mechanisms of functional materials. Particular attention is paid to the drug-receptor binding process and to the interpretation of the enthalpy-entropy compensation phenomenon.
Article
A simple model of the charge distribution in a π-system is used to explain the strong geometrical requirements for interactions between aromatic molecules. The key feature of the model is that it considers the σ-framework and the π-electrons separately and demonstrates that net favorable π-π interactions are actually the result of π-σ attractions that overcome π-π repulsions. The calculations correlate with observations made on porphyrin π-π interactions both in solution and in the crystalline state. By using an idealized π-atom, some general rules for predicting the geometry of favorable π-π interactions are derived. In particular a favorable offset or slipped geometry is predicted. These rules successfully predict the geometry of intermolecular interactions in the crystal structures of aromatic molecules and rationalize a range of host-guest phenomena. The theory demonstrates that the electron donor-acceptor (EDA) concept can be misleading: it is the properties of the atoms at the points of intermolecular contact rather than the overall molecular properties which are important.
Book
Helps you choose the right computational tools and techniques to meet your drug design goals Computational Drug Design covers all of the major computational drug design techniques in use today, focusing on the process that pharmaceutical chemists employ to design a new drug molecule. The discussions of which computational tools to use and when and how to use them are all based on typical pharmaceutical industry drug design processes. Following an introduction, the book is divided into three parts: Part One, The Drug Design Process, sets forth a variety of design processes suitable for a number of different drug development scenarios and drug targets. The author demonstrates how computational techniques are typically used during the design process, helping readers choose the best computational tools to meet their goals. Part Two, Computational Tools and Techniques, offers a series of chapters, each one dedicated to a single computational technique. Readers discover the strengths and weaknesses of each technique. Moreover, the book tabulates comparative accuracy studies, giving readers an unbiased comparison of all the available techniques. Part Three, Related Topics, addresses new, emerging, and complementary technologies, including bioinformatics, simulations at the cellular and organ level, synthesis route prediction, proteomics, and prodrug approaches. The book's accompanying CD-ROM, a special feature, offers graphics of the molecular structures and dynamic reactions discussed in the book as well as demos from computational drug design software companies. Computational Drug Design is ideal for both students and professionals in drug design, helping them choose and take full advantage of the best computational tools available.
Article
Crystallographic and theoretical (ab initio) data on intermolecular nonbonded interactions have been gathered together in a computerised library (`IsoStar'). The library contains information about the nonbonded contacts formed by some 250 chemical groupings. The data can be displayed visually and used to aid protein–ligand docking or the identification of bioisosteric replacements. Data from the library show that there is great variability in the geometrical preferences of different types of hydrogen bonds, although in general there is a tendency for H-bonds to form along lone-pair directions. The H-bond acceptor abilities of oxygen and sulphur atoms are highly dependent on intramolecular environments. The nonbonded contacts formed by many hydrophobic groups show surprisingly strong directional preferences. Many unusual nonbonded interactions are to be found in the library and are of potential value for designing novel biologically active molecules.