Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5319_Библиотеки_им_академика_М_И_Перельмана.pdf
X
- •Preface and Acknowledgement
- •Chemical Structures of Amino Acids,Molecular Graphics and Introduction
- •Introduction
- •Literature
- •Chapter Abstract Videos
- •Contents
- •About the author
- •1.10 Synopsis
- •1.3 The Battle Against Infectious Disease
- •1.4 Biological Concepts in Drug Research
- •Bibliography and Further Reading
- •2.8 A Long List of Accidents
- •2.10 Synopsis
- •Bibliography and Further Reading
- •3. Classical Drug Research
- •3.2 Malaria: Success and Failure
- •3.6 Synopsis
- •Bibliography and Further Reading
- •4.1 The Lock-and-Key Principle
- •4.2 The Essential Role of the Membrane
- •4.6 Blame It All on Water!
- •4.11 Lessons for Drug Design
- •4.12 Synopsis
- •Bibliography and Further Reading
- •5.1 Louis Pasteur Sorts Crystals
- •5.2 Structural Basis of Optical Activity
- •5.4 Lipases Separate Racemates
- •5.8 Synopsis
- •Bibliography and Further Reading
- •6.2 Lead Structures from Plants
- •6.9 Synopsis
- •Bibliography and Further Reading
- •7.2 Color Change Demonstrates Activity
- •7.7 Biophysics Supports Screening
- •7.11 Synopsis
- •Bibliography and Further Reading
- •8.1 Strategies for Drug Optimization
- •8.5 From Agonists to Antagonists
- •8.9 Synopsis
- •Bibliography and Further Reading
- •9. Designing Prodrugs
- •9.1 Foundations of Drug Metabolism
- •9.2 Esters Are Ideal Prodrugs
- •9.6 Synopsis
- •Bibliography and Further Reading
- •10. Peptidomimetics
- •10.1 Therapeutic Relevance of Peptides
- •10.2 Designing Peptidomimetics
- •Bibliography and Further Reading
- •11.4 What Is Contained in Chemical Space?
- •Bibliography and Further Reading
- •12.7 Silencing Genes by RNA Interference
- •12.9 Proteomics and Metabolomics
- •Bibliography and Further Reading
- •13.3 Crystal Lattices Diffract X-Rays
- •Bibliography and Further Reading
- •Bibliography and further reading
- •15. Molecular Modeling
- •15.2 Strategies in Molecular Modeling
- •15.3 Knowledge-Based Approaches
- •15.4 Force Field Methods
- •15.5 Quantum Chemical Methods
- •Bibliography and further reading
- •16. Conformational Analysis
- •16.8 Synopsis
- •Bibliography and Further Reading
- •Bibliography and Further Reading
- •18.4 Lipophilicity and Biological Activity
- •Bibliography and Further Reading
- •19.3 The Role of Hydrogen Bonds
- •19.5 Absorption Profiles of Acids and Bases
- •19.8 From In Vitro to In Vivo Activity
- •Bibliography and Further Reading
- •Bibliography and Further Reading
- •21.5 LUDI Discovers the First Leads
- •Bibliography and Original Papers
- •22.1 The Druggable Genome
- •22.4 Enzymes and Their Inhibitors
- •22.9 Resistance and Its Origin
- •Bibliography and Further Reading
- •23.1 Serine-Dependent Hydrolases
- •23.10 Synopsis
- •Bibliography and Further Reading
- •24. Aspartic Protease Inhibitors
- •24.2 Design of Renin Inhibitors
- •24.8 Synopsis
- •Bibliography and Further Reading
- •25.1 Structure of Zinc Metalloproteases
- •25.9 What Zinc Can Do, Iron Can Too
- •25.11 Synopsis
- •Bibliography and Further Reading
- •26. Transferase Inhibitors
- •26.1 The Kinase “Gold Rush”
- •Bibliography and Further Reading
- •27. Oxidoreductase Inhibitors

. • The Pharmacophore is Modied by Conformational Transitions
. Fig. 17.4 Inhibitors of the angiotensin-converting enzyme.
A pharmacophore that consists of a terminal carboxylate group,
acarbonyl group, and agroup that coordinates to the catalytic zinc
ion is necessary for binding to the enzyme. The latter function is as-
(. Fig.17.5) are considered. They are ligands of an enzyme in ergosterol biosynthesis. Spring forces are applied
between the atoms marked with the same numbers. The
minimization of these forces together with the individual
force elds of the four molecules leads to the superposition shown in . Fig.17.5.
sumed by athiol, aphosphoric, phosphonic, or carboxylic acid group.
The individual derivatives possess conformational exibility in different areas
Unfortunately, the resulting solution depends on the
starting conditions. If the molecules are oriented differently in space at the beginning of the calculation, or if
they start from different conformations, different superpositions can result. At rst glance, this argument may
seem rather implausible. It should be kept in mind that
molecules are not only considered under the inuence

1
Chapter • Pharmacophore Hypotheses and Molecular Comparisons
17
. Fig. 17.5 “Virtual” springs are coupled to the atoms that are
marked with numbers around the steroid 17.2 and the three derivatives 17.3–17.5. The structural superposition (bottom) that is shown is
of “virtual” spring forces, but also under their own force
elds. The multiple minima problem encountered in molecular force eld calculations was already mentioned in
Chap.16. It also plays an important role here. The hiker
from the last chapter should help to explain this issue. He
stands on amountaintop and wants to descend into the
deepest valley possible. At the same time, he feels an “extra force” because he is very thirsty. He wants to meet his
friends in apub. The friends come from different peaks
in the mountains. He sees apub in each valley. But which
one should he choose? For acommon meeting place, he
would also accept aless deep valley. At the beginning of
his hike, he looks for the steepest descent to get down
quickly. After awhile, the other valleys fall out of sight.
When he nally arrives at another pub, he has no longer the energy to look for adifferent one. If he would
have started from adifferent peak, he might have found
asimilarly deep valley, but he would have found the pub
of his choice and met his friends at the same time. The
problem of choosing starting conditions for molecular
comparisons with “virtual” spring forces is similar. How
can you verify that the best possible solution has been
found? Only an experiment can help. This requires the
synthesis of molecules that are conformationally rigid in
certain regions due to the insertion of rings. They impart
axed spatial arrangement of the pharmacophore. If
determined by the forces of these springs and the simultaneous consideration of the individual molecular force elds
these molecules also have activity, their rigid geometry
indicates acorrect selection of the pharmacophore (see
Sect.17.9).
17.5 Systematic Conformational Search
and Pharmacophore Hypothesis:
The “Active Analog Approach”
The last chapter focused on conformational analysis.
Could the techniques described there, such as the systematic rotation around certain bonds, be used to search
for the spatial arrangement of the pharmacophore? Garland Marshall at Washington University, St. Louis, USA,
developed such atechnique, called the “active analog ap-
proach,” in the late 1970s. First, putative pharmacophoric
groups must be assigned to all molecules in adataset.
Then, the expected equivalence of the groups must be
dened, meaning which groups are equivalent to which
other groups. Asystematic conformational search is then
performed on the rst compound in the dataset. During
the search, the distances between the functional groups
of the putative pharmacophore are determined for each
specied geometry. These distances are stored. Since
molecules cannot adopt just any geometry, the distances
will fall into certain intervals. The same procedure is used

. • Molecular Recognition Properties and the Similarity of Molecules
for the second molecule in the dataset. In principle, only
the distance ranges specied by the rst molecule need to
be searched. It may be that all the distances found with
the second molecule were already found with the rst. It
is also possible that certain distances are excluded by the
second molecule, thus, narrowing the “allowed” distance
ranges. All molecules in the dataset are analyzed one after the other in this way.
If the conformational exibility of the molecules is
restricted in different regions of their scaffolds, there will
be achance that only one or afew distance ranges remain
for the functional groups of the pharmaco phore. This
leads to the possible binding geometries of the pharmacophore groups in the ligands. Finally, geometry optimization can be performed, for which the “virtual” spring
method is appropriate. This method is ideal at this stage
because the desired solution is already very close. It is
easy to imagine that the order in which the molecules
are examined is crucial for the efciency of the “active
analog” procedure. It is best to start with the most rigid
molecule in the dataset. With alittle luck, this will al
ready restrict alarge part of the accessible conformational space. The list of possible distances will then remain rather small. By consistently using such constraints,
Garland Marshall and his research group were able to
propose in 1987 amodel for the receptor-bound conformation of the ACE inhibitors shown in . Fig.17.4.
What could be more rewarding than being able to personally validate this model years later and nd that it was
correct within asurprisingly small margin of error! The
validation became possible because the crystal structure
of the enzyme with several inhibitors from this dataset
had been successfully determined in 2003 (see Sect.25.5).
17.6 Molecular Recognition Properties
-
and the Similarity of Molecules
The question is whether the concepts presented in the
previous sections for representing the properties of
molecules have really been adequately considered in the
comparisons attempted. It is not easy to decide which
functional groups correspond to the individual “anchor
points” of a pharmacophore. Analogous functional
groups must be oriented in asimilar spatial direction
in all molecules. In the case of the ACE inhibitors
(. Fig.17.4), conicts arise even in the assignment of
functional groups. Some analogues carry two carboxylate groups, which must be unambiguously assigned to
the pharmacophore before they can be compared with
other inhibitors.
The binding of small-molecule ligands to aprotein
is amutual, targeted recognition process. Both partners
must t together to form astrong interaction. Parts of
the ligand with complementary recognition properties
determine the binding to the receptor. The term “recog-
nition properties” refers to all properties that contribute
to the specic interaction between molecules. Until now,
only those properties and similarities that can be read
directly from the molecular scaffold have been considered.
But is that enough? What would the world be like if we
could only recognize each other by our bones? Not even
male and female would be immediately distinguishable!
All the stimuli of interpersonal relationships conveyed
by personal appearance and charisma would be lost. So
far, molecules have been considered only in terms of their
“molecular skeletons.” Why should it be possible to describe ligand–receptor interactions at this level? Molecules also recognize each other by their shape and surface,
as well as by the properties they transmit to their immediate environment and use to make contacts. The following
example should illustrate this point. Methotrexate 17.6
(MTX) and dihydrofolate 17.7 (DHF) bind to the enzyme
dihydrofolate reductase (. Fig.17.6 andSect.27.2). The
side chains of the two molecules are almost identical, but
the heterocyclic moieties are different. It is known from
NMR spectroscopic studies that the protonated form
of MTX binds to the protein. Considering the chemical
formulas, it is tempting to superimpose the two heterocycles directly. Perfect scaffold equivalence is achieved, and
the heteroatoms in both molecules fall on top of each
other. The receptor, however, does not care about an
apparent equivalence of the molecular skeletons. Much
more important is the interaction with the surface of the
molecule. Polar molecules such as MTX or DHF will be
bound to the protein by hydrogen bonds. The arrows in
. Fig.17.6 characterize the H-bond donor and accep-
tor groups. The arrows point towards the molecule when
an acceptor site is exposed and away from the molecule
when adonor site is exposed. Initially, the molecules are
oriented in space so that they correspond with respect to
adirect atom–atom matching. For the moment, the basic
molecular skeleton should be ignored and only the distribution of H-bond donor and acceptor groups should
be considered (. Fig.17.6, bottom left). The obtained
equivalence is not very convincing. Therefore, another
variant is considered in which the heterocycle of DHF is
ipped along the bond between the heterocycle and the
side chain. The spatial overlap of the two molecules is
no longer optimal, but the pattern of exposed donor and
acceptor groups for both molecules now shows amuch
better match (. Fig.17.6, bottom right). In adifferent
conformation, the molecules now present themselves with
altered but much better matching molecular recognition
properties. Even atrained eye can hardly read these differences from the structural formulas alone, even in acase
as clear as the one presented here.
Models are ne, but are they actually correct? This
can only be answered by experiment. Fortunately, in
the present case, crystal structures are available for
both ligands complexed with dihydrofolate reductase
(DHFR). The observed binding geometries are shown

1
Chapter • Pharmacophore Hypotheses and Molecular Comparisons
17
. Fig. 17.6 Methotrexate 17.6 and dihydrofolate 17.7 are ligands of
dihydrofolate reductase. The side chainR (see Sect.27.2, . Fig.27.9)
is identical for both except for a methyl group on the nitrogen atom.
The heterocycles are different. Upper row Intuitively, superposition of
both heterocycles directly upon each other when comparing the structures appear reasonable. Heteroatoms match pair-wise to one another.
Bottom row Arrows are distributed around the molecules to compare
the hydrogen-bonding properties. They are pointed to the molecule
in . Fig.17.7. One aspartate and two carbonyl groups
in the main chain and two water molecules are responsible for recognition in the binding pocket. The water
molecules mediate hydrogen bonds between the ligand
and the protein. The experimentally determined binding
geometries show that the described considerations about
the similarity of the hydrogen bonding properties led to
the right conclusions. Asurprising and apparently “nonequivalent” orientation of the two ligands in the binding
pocket is easily explained. The properties responsible for
the mutual recognition process must be compared. These
are the only features that count in the comparison! It is
noteworthy that the experimental conrmation of the
considerations described above came eight years after the
working hypothesis was proposed. This is anice example
of the performance of the model hypothesis.
Other properties besides hydrogen bonding can serve
as additional criteria to dene similarities in the molecular recognition process. The electrostatic potential
(Chap.15) calculated for the heterocyclic ring systems
of DHF and MTX (. Fig.17.7) suggests very similar
when an acceptor (red) is present and they point away for donor
groups (green). If the molecular skeletons are masked out, and the distribution of H-bond donor and acceptor groups is concentrated upon,
the atom–atom overlap obtained via the direct superposition of the
rings will show rather unconvincing equivalence (left). Instead, if the
heterocycle in 17.7 is ipped about the bond between the heterocycle
and the side chainR, the pattern of donor and acceptor groups that
will be obtained now exhibits convincing equivalence (right)
conclusions. In addition to the aforementioned H-bonding properties and electrostatic potentials, steric space
lling and the distribution of hydrophobic properties on
the surface of both ligands play an important role. When
molecules are superimposed to predict their putative geometries in the binding pocket, their conformational exibility must also be taken into account.
17.7 Automated Molecular Comparisons
and Superpositioning Based on
Recognition Properties
Is it possible to consider all of the properties mentioned in the last section in amethod for superimposing molecules for relative comparison? To accomplish
this, ameasure of similarity must be computed for all
the properties. This measure must be related to aspatial
distance function. Then the spatial superposition can be
performed. At the same time, the maximum similarity
of the selected features is sought. The program SEAL

. • Automated Molecular Comparisons and Superpositioning Based on Recognition Properties
. Fig. 17.7 Experimentally determined binding geometries of meth-
otrexate (17.6, gray carbon atoms) and dihydrofolate (17.7, green carbon atoms) in dihydrofolate reductase. The heterocycles of the ligands
are bound through H-bonds to the carboxylate or carbonyl group of
an amino acid that is oriented into the binding pocket. Two water molecules (red spheres) mediate additional H-bonds between the ligands
and the protein. The difference in the binding modes that is discussed
in . Fig.17.6 is well approved. On the right-hand side, the electro-
static potentials around methotrexate (top) and dihydrofolate (bottom)
by Simon Kearsley and Graham Smith at Merck Sharp
& Dohme determines the spatial similarity of different
properties distributed over the molecular scaffold. It
simultaneously ranks the similarity with respect to the
overlap volume of the molecules determined during superposition. In this way, the superposition of MTX and
DHF is correctly predicted according to the experiment.
The conformational exibility is also considered in this
analysis. For this purpose, precalculated conformers can
be taken and successively compared with one another.
This is implemented in the program ROCS by Anthony
Nicholls at OpenEye, USA. Christian Lemmen at GMD
in St. Augustin, Germany, has taken aslightly different
approach in the program FlexS. First, areference ligand
is represented by aset of property-loaded Gaussian functions. In effect, the molecule is described as a density
distribution of pharmacophore properties in space. The
candidate molecule to be superimposed on the reference
molecule is then taken. It is broken down into fragments.
First, acentral base fragment is placed on the reference
so that its description by Gaussian functions overlaps
the reference as much as possible. Then the other frag-
are shown. The molecules are shown in aspatial orientation that was
determined by crystal structure analysis. Considered qualitatively, the
electrostatic potentials of both molecules have very similar spatial
course in this orientation. (7 https://sn.pub/UxHUJF)
ments are added to the base fragment until the complete
ligand is recovered at the end. During this incremental
attachment procedure, care is taken to ensure that the
fragments added to the base fragment also t optimally
into the Gaussian description of the reference. At the
same time, the conformational exibility of aligand is
taken into account when adding subsequent fragments.
Acomplication arises when analyzing the similarity
of molecules using these methods. It is assumed that
the relevant properties dening the similarity have been
found. However, the question arises as to what is accepted as “sufciently” similar for acomparable effect at
areceptor. There is atoy called a“shape sorter” in which
children try to t blocks of different shapes into abox
through holes which are prepunched for them. There
is amatching hole for each block shape, cube, cuboid,
and cylinder with acircular or elliptical cross section.
In asimilarity analysis, one is tempted to group cubes
and cuboids, or “circular” and “elliptical” cylinders, as
related because of their shape. When you try to t the
pieces through the holes in the box, you may nd that
the cube ts not only through the rectangular hole, but

Chapter • Pharmacophore Hypotheses and Molecular Comparisons
1
17
also, perhaps with some effort, through the hole for the
elliptical cylinder. The cube is only slightly too large to
t through the square hole and the hole for the circular
cylinder. So are the cube and the elliptical cylinder or the
cube and the circular cylinder more similar? The measure
of similarity applied to the molecules is calibrated to the
receptor into which the molecules are supposed to t.
Thus, it is always arelative measure!
Thiorphan and retro-thiorphan (Sect.5.5, formulas
5.23 and 5.24) differ only in the spatial sequence of the
amide bond. They bind with almost identical afnity to
the zinc protease thermolysin, and NEP 24.11. Therefore, they would be classied as very similar. The zinc
protease ACE binds thiorphan by at least afactor of
100 times more strongly than retro-thiorphan (Sect.5.5,
. Fig.5.12). Relative to this enzyme, both substances
must be considered dissimilar. Another extreme is seen
in the oligopeptide-binding proteinA (Sect.4.1). It binds
any tri- to pentapeptide containing acentral Lys–Xxx–Lys
unit (Xxx: any amino acid) with almost the same afnity. In principle, only information about the shape of the
binding site is needed for asimilarity analysis. Only then
can the requirements be adequately dened. However, in
adrug design project, the structure of the receptor may
still be unknown; thus there is no choice: only hypotheses
and their experimental testing in incremental steps can
approximate the structural requirements of the receptor.
17.8 Rigid Analogues Trace the
Biologically Active Conformation
The concepts in Chap.16 showed that an enormously
large number of conformers can be easily generated for
many drug-like molecules. If acomparison of all con
formers is desired, the undertaking will quickly become
computationally intensive. How can we obtain arelevant
image of the bound conformations? Either acompound
in the dataset is highly rigid and constrains the putative arrangements of the pharmacophore in space, or
the molecules under consideration are rigid in different
regions of their molecular scaffold. . Fig.17.8 shows
the structural superposition of the steroid 17.2 with the
inhibitors 17.3–17.5 described above. This result was
obtained from asimilarity analysis with multiple conformers and the superposition is very similar to the calculation with the “virtual” spring forces. It has, however,
adecisive advantage: we do not need preconceived denitions of equivalent centers between which the spring
forces are applied. These equivalences arise automatically
through asimilarity comparison of the properties that
are distributed over the molecules.
17.9 If Rigid Analogues are Lacking: Model
Compounds Elucidate the Active
Conformation
In the last example, alargely rigid reference compound was
provided. How to proceed if no such reference compound
is known? Only experiment can help. Rigid analogues have
to be synthesized. These are tested for biological activity. If
they still show afnity for the receptor, it can be assumed
that the active conformation has been frozen.
The following example illustrates how the receptor-bound conformation can be studied by synthesizing
rigid model compounds. The calcium channel blocker
nifedipine 17.8 (Sects.2.5 and30.4) contains several rotatable bonds (. Fig.17.9). It can, therefore, adopt many
conformations. What is the orientation of the phenyl ring
relative to the dihydropyridine ring? This question was
elegantly answered by Wolfgang Seidel at Bayer through
the synthesis and crystal structure determination of the
cyclized derivatives 17.9. An additional lactone ring
changes the biological activity of the derivative depending on the ring size. In compounds with asix-membered
-
lactone, the phenyl and dihydropyridine rings are almost
in the same plane. Conversely, in the twelve-membered
ring derivative, the phenyl ring is perpendicular to the
dihydropyridine ring. The afnity of this compound
is about ve orders of magnitude higher than that of
the six-membered lactone derivative. Therefore, it was
hypothesized that nifedipine acts in aconformation in
which the phenyl and dihydropyridine rings are perpen-
. Fig. 17.8 Superposition of the steroid 17.2 and three inhibitors 17.3–17.5 according to aspatial comparison of their molecular properties. In
contrast to methods with “virtual” spring forces, this method does not require apredened equivalence of molecular groups. It is automatically
generated by the similarity comparison of many different conformations

. • The Protein Denes the Pharmacophore: “Hot Spot” Analysis of the Binding Pocket
dicular to each other. Many years later, this hypothesis
was shown to be correct. Acryo-EM structure of the
calcium channel with bound nifedipine demonstrates the
perpendicularity (Sect.30.4).
After this question has been answered, more compounds can be designed. Arelevant superposition that
corresponds to the conditions in the protein’s binding
pocket will be possible. Such superpositions have gained
adecisive meaning in the context of 3D structure–activity relationships (Chap.18).
17.10 The Protein Defines the
Pharmacophore: “Hot Spot”
Analysis of the Binding Pocket
It was described in Sect.17.1 that apharmacophore can
also be deduced from the protein structure. The com-
puter program GRID from Peter Goodford is atool that
is often used for this purpose. It calculates favorable positions in protein-binding pockets for functional groups
of apotential ligand. This could be acarboxylate group,
ahydroxy group, or an aliphatic carbon atom. The potential function used for GRID has been calibrated for
avariety of functional groups on the crystal structures
of organic molecules. The result of aGRID calculation
is aset of interaction energies at each point of intersection of agrid that is inscribed in the binding pocket.
The energies are presented graphically, for example, by
indicating the region of space where the interaction energy meets or exceeds apredened threshold. They indicate hot spots for the placement of functional groups of
apotential ligand. The areas in which the interactions
with an aromatic carbon atom or ahydroxyl oxygen atom
are favorable are shown for the enzyme thermolysin in
. Fig.17.10. Such calculations are carried out with aset
of different probes, for instance, awater molecule, an
aromatic carbon, ahydrogen-bond acceptor or donor,
. Fig. 17.9 The calcium channel blocker nifedipine 17.8 contains
multiple rotatable bonds. The phenyl ring can coincide with aplane
of the dihydropyridine ring or they orient perpendicular to each other. To distinguish between these possibilities, lactones with different
ring size 17.9 were synthesized and their crystal structures were determined. The phenyl ring lies almost parallel to the dihydropyridine
ring (α ≈ 0°) in the compound with the six-membered-ring lactone
(orange). Upon increasing the ring size, the angle between the two
rings grows so that aperpendicular orientation (α ≈ 80°) is achieved
in the twelve-membered-ring derivative (green). The biological activ-
ity increases from virtually inactive, as in the six-membered ring, to
almost ve orders of magnitude higher for the twelve-membered-ring
derivative. The bioactive conformation of nifedipine (gray), therefore,
requires aperpendicular orientation of the two rings.
(7 https://sn.pub/tNnF5T)

1
Chapter • Pharmacophore Hypotheses and Molecular Comparisons
17
. Fig. 17.10 An analysis of the binding pocket of thermolysin. Ar-
eas of favorable interactions were calculated for an aromatic carbon
probe (white) and ahydroxyl oxygen atom (red). There are also fragments mentioned in . Fig.7.10 that could be determined by allowing
the probe molecules to diffuse into the protein crystals. The calculated
or apositively or negatively charged group. The results
provide valuable information about the shape and electrostatic properties of the binding pocket.
Another approach to protein structure analysis is
based on the idea that the physical nature of nonbinding
interactions is identical in protein–ligand complexes and
in the crystal packing of small organic molecules. The latter are particularly interesting for this purpose because the
crystal structures of small organic molecules are regularly
determined with great precision. More than 1.25 million
crystal structures are stored in the Cambridge Database
(Sect.13.9). This collection is ideal for obtaining relevant
and reliable data for ligand design through statistical analysis of crystal packing data (Sect.14.7). Suppose there is
hot spot corresponds well with the positions that were crystallographically determined with molecular probes. (7 https://sn.pub/m3N94q)
acarboxylate group –COO– on the protein that protrudes
into the binding pocket. Where must apartner group be
positioned to form afavorable interaction? To answer this
question, the Cambridge Database was rst searched for
compounds containing carboxylate groups. Then, for each
of the retrieved acid groups, the position of the counter
group forming an H-bond to the carboxylate was stored.
Finally, the collective of all H-bonds found was superimposed by exactly mapping the carboxylate groups of all examples to one another. The distribution of H-bond donor
groups (. Fig.17.11) provides acomposite picture of the
allowed range of H-bonding geometries. Such adistribution can then be superimposed on the protein structure by
matching it to the carboxylate group of the residue in the

. • The Protein Denes the Pharmacophore: “Hot Spot” Analysis of the Binding Pocket
. Fig. 17.11 Hydrogen-bonding geometries (carbon is green, oxygen
is red, and hydrogen is white) around acarboxylate group(a), ester
group(b), carbonyl group(c), and ether group(d). Structures with
these central groups that form hydrogen bonds with OH-donor groups
were extracted from the Cambridge Structural Database. These examples were superimposed based on the geometry of the central group.
protein. Areas where the distribution overlaps with other
atoms in the protein are discarded. In this way, the energetically most favorable sites for acounter group in the
binding pocket are identied. . Fig.17.12 compares these
distributions with aprotein–ligand complex. As expected,
the hydrogen bond geometries found in the complex agree
well with the range found in the crystal packing of organic
molecules. Asystem of rules for nonbonding interactions
in protein–ligand complexes was obtained from the statistical evaluations of all the groups found in proteins. These
rules are compiled at the Cambridge Crystallographic
Data Centre in the Isostar database. Once overlaid on the
It is obvious that there is considerable variability in the interaction
geometries, but also that preferred orientations are to be found. It is
also shown that, for instance, the interaction pattern around an ester
group(b) is not simply an additive superposition of the distribution
around acarbonyl group(c) and an ether group(d)
protein, they can be contoured to map binding hotspots
using the SuperStar program.
Knowledge-based potentials are another approach to
represent aprotein-based pharmacophore. For this, the
contact geometries in protein–ligand complexes are evaluated. Ahistographical distribution is generated showing
how often aparticular contact occurs between agroup
found in aligand and in an amino acid of a protein.
When such astatistical frequency distribution is related
to amean reference state, an energy function can be calculated from it. This function assumes that contacts, occurring more frequently than the average distribution, are

1
Chapter • Pharmacophore Hypotheses and Molecular Comparisons
17
. Fig. 17.12 The distribution of H-bond donor groups (carbon
is white, oxygen is red, and nitrogen is blue) around a carboxylate
group or acarbonyl group are superimposed with the 3D structure
of the complex of methotrexate with dihydrofolate reductase (cf.
. Fig.17.7). The distributions are imposed onto the acid group of
Asp26 and the carbonyl groups of Leu 4 and Ala97. The hydrogen
bonds formed between protein and ligand coincide geometrically with
energetically favorable. If they occur less frequently than
the average, they will be considered unfavorable. These
statistical potentials have been incorporated into the scoring function DrugScore, developed by Holger Gohlke in
Marburg, Germany. They can also be used for graphical
analysis of binding pockets and help to indicate the spatial distribution of hot spots for ligand binding.
The MCSS method was developed in the group of
Martin Karplus at Harvard Medical School, Boston,
USA. Several thousand random probe molecules such
as acetone, water, methanol, or benzene were placed
in abinding pocket for this. Acomputer simulation is
started with which the single probe molecules are moved
into optimal positions. They are driven by acalculation
according to the underlying force eld. The probe molecules experience the interaction with the protein, but they
do not “see” one another. At the end of the calculation,
afrequency distribution for the probe molecules is obtained. If this distribution is evaluated, ahot spot for an
interaction with the protein can be highlighted. If the
thus obtained hot spots are compiled into acomposite
picture, aprotein-based pharmacophore will be obtained.
ranges often found in crystal packings of small organic molecules.
(7 https://sn.pub/rpQuWI)
17.11 Searching for Pharmacophore
Patterns in Databases Generates
Ideas for Novel Lead Compounds
Apharmacophore can be used to search adatabase for
promising candidates that can be accommodated in the
binding pocket of a protein. The reference pharmacophore can either be derived from aset of superimposed
ligands, or areference protein can dene its properties.
How such adatabase search is performed and what is
discovered depends on how much information is stored
in the database itself. If only 2D structures are collected,
all examples with acertain functional group or substructure can be retrieved. Based on the topology, different
criteria are dened to determine the degree of similarity
between molecules. If the denition of the pharmacophore is very general, e.g., an aromatic compound with an
acidic group and abasic nitrogen atom, many hits will
be found. However, the relative spatial distances between
these groups are important. Such information is not considered when searching a2D database. Matthias Rarey
and Scott Dixon at the time at Smith Kline & Beecham
Соседние файлы в папке Библиотека им академика М.И. Перельмана
