1. Pharmacophore
in Drug Design

2. Abstract
Drug Design’s goal: to develop new ligands with high binding affinity toward a protein receptor.
Pharmacophore: 3D arrangement of essential features that enable a molecule to exert a particular biological effect.
A major goal in contemporary drug design is to develop new ligands with high affinity of binding toward a given protein receptor.
Pharmacophore, which is the three-dimensional arrangement of essential features that enable a molecule to exert a particular biological effect, is a very useful model for achieving this goal.

3. Computer-Aided Drug Design
Is receptor structure
available No
Yes
The two important tools involved in predicting molecular-interactions in computer aided drug design (CADD) are pharmacophore-based and docking techniques.
If the 3D structure of the receptor is known, pharmacophore is a complementary tool to standard techniques, such as docking.
However, frequently the structure of the receptor protein is unknown and only a set of ligands together with their measured binding afiinities towards the receptor is available. In such a case, a pharmacophore-based strategy is one of the few applicable tools.
Pharmacophore validation through docking
Docking
Pharmacophore
Pharmacophore as a constraint in docking

4. Agenda Definition Computer-Aided Drug Design Flow
Pharmacophore Identification
Indirect methods
Direct methods
Pharmacophore fingerprints
Applications for Drug Design
Pharmacophore fingerprints are 1D descriptors that encode pharmacophore information.

5. What is Pharmacophore? Paul Ehrlich (~1900): Later:
The molecular framework that carries (phoros) the essential features responsible for a drug's (pharmacon) biological activity.
Later:
it became clear that the 3D disposition of the pharmacophoric features is also important.
The term pharmacophore was introduced by Paul Ehrlich in the early 1900s, to refer to the molecular framework that carries (phoros) the essential features responsible for a drug's (pharmacon)
biological activity" .
Later, as molecular structures were worked out, it became clear that the presence of pharmacophoric features was insufficient for activity and their disposition in 3Dl space was also critically important.
Presently, the term has been expanded to refer to the 3D arrangement of features that enable a molecule to exhibit a specific biological activity [3]. In other words, molecules are active in a particular receptor if they carry a number of features that interact favorably with the receptor and which possess a geometry complementary to it

6. The figure shows the importance of a pharmacophore in drug design and presents the main issues that will be discussed in the current review.
It illustrates that there are two different ways to deduce a pharmacophore: direct methods and indirect methods. Direct methods use both ligand and receptor information, whereas indirect methods use only a collection of ligands that have been experimentally observed to interact with a given receptor. Frequently, the exact structure of the receptor is unknown. In such a situation, only indirect methods are applicable. However, direct methods are becoming extremely important with the rapidly increasing number of known protein structures, which is the outcome of the Structural Genomics project.
Once identified, a pharmacophore model can be a versatile tool to aid in the discovery and development of new lead compounds. The power of pharmacophore-based methods for lead generation lies in their ability to suggest a diverse set of compounds potentially possessing a desired biological activity, but which have totally different chemical scaffolds. This is the main advantage of pharmacophore-based method over lead explosion, which usually generates a set of molecules that are very similar to the original lead. If a drug candidate fails for ADME/Tox reasons, similar candidates may also fail. Therefore, a diverse set of leads increases the changes that some of compounds will pass all the stages of the drug development process. A pharmacophore model can be used to discover new active ligands by using it as a query to search a database of 3D chemical structures. Alternatively, it can be used in de novo design to develop totally novel drugs that satisfy the pharmacophore requirements. Furthermore, a pharmacophore model can guide chemists to synthesize new compounds during lead optimization. Potential drug leads identified by in-silico methods must be further verified by experimental tests. The results of the experimental tests provide useful information for the improvement of the pharmacophore model.

7. Indirect Methods for Pharmacophore Identification

8. Flow Conformational Search Feature Extraction Structure Representation
Ligands & their affinity
Conformational Search
Feature Extraction
Structure Representation
The following are the different stages for identifying a pharmacophore pattern from a set of ligands that interact with the same receptor:
1. Input - The ligand data set for which the pharmacophore model will be constructed is selected.
2. Conformational Search - Addressing the flexibility of the input ligand can be done either as a separate initial step or combined within the pattern identification process (stage 5). Sometimes when there is a sufficient data set of crystal structures that cover the conformational space, this stage can be omitted.
3. Feature Extraction - The features to be used in the pharmacophore pattern discovery are extracted from the input ligands. These features can be atoms (e.g.
carbon), topological groups (e.g. phenyl ring) or functional groups (e.g. hydrogen bond donor).
4. Structure Representation - The features of each ligand are combined to represent its structure.
5. Pattern Identification - The features extracted from different ligand molecules are matched and pharmacophore candidates are proposed.
6. Scoring - The pharmacophore candidates are ranked and the highest-scoring ones are reported to the user.
Pattern Identification
Pharmacophore Candidates
Scoring

9. Input – Dataset of ligands
Ligand type
Usually active and share the same activity
Sometimes inactive ligands are used
Rarely different activity levels are used
Ligand Diversity
Dataset size
Usually dataset size < 100
- Ligand Type: In most of the methods all the input ligands are considered to be active. Furthermore, these methods usually assume that the ligands share the same activity and thus they do not consider activity data. In other methods information about inactive ligands is also used. This information is highly important to indicate the structural features that significantly decrease the activities. When different activity levels are available for a large enough set of diverse compounds, this information can be utilized to derive a pharmacophore model that can be used to predict the activity of unknown compounds.
Data Set Diversity: In order to get an accurate pharmacophore model, the data set should be as diverse as possible. This will allow to identify features that are most critical for the binding. However, it is important that the outliers will not have a high influence on the obtained model. In addition, one should remember that very different ligands may bind at different binding sites and this may lead to a wrong pharmacophore model.
- Data Set Size: Most of the current available methods are designed to handle small data sets, which are composed of less than 100 ligands. Other methods get as an input a large data set of ligands, but they convert it into a smaller one by sorting the activities of the ligands and then considering only a small number of the most active ones. The drawback of this strategy is that it ignores much of the data and is highly influenced by the cutoff value.

10. Conformational Search
The pharmacophore identification problem is complicated substantially by the fact that ligands are very flexible molecules. That is, they possess many internal degrees of freedom. As a result, a ligand may have many possible conformations. Each conformation may bind in the active site of the considered receptor. Thus, all the conformations of each input ligand have to be considered during a search for a pharmacophore.

11. Conformational Search (Cont.)
Separated
Initial Stage
Combined within
the process
Separated initial stage: Most of the current existing pharmacophore identification methods treat the conformational search as a separate initial stage. Specifically, a set of conformers is generated with the goal of covering the whole conformational space of each ligand. Then, pharmacophores are identified from this set of conformers. The main advantage of this approach is its modularity i.e. a variety of conformational searching tools can be used. The main drawback of this approach is that the number of conformations required to cover the whole conformational space is extremely large, especially for highly flexible compounds. Thus, there is a tradeoff between the size of the representative conformational set and the amount of conformational space that is covered for each compound. Their are two major ways to handle this problem:
Energy Minimization: In order to reduce the number of representative conformers, it is often assumed that only low-energy conformations are stable. Based on this assumption, energy minimization techniques are used to restrict the generated conformations within the low-energy region. However, it should be noted that this assumption is not always correct.
Redundancy Elimination: In order to eliminate redundant conformations, the conformational search method is followed by a clustering analysis or is combined with some method that is specifically designed for ensuring conformational diversity. An example for a method that ensures conformational diversity is the poling technique, which penalizes a newly generated conformation if it is too close to an already existing one.
Combined within the process: An alternative approach is to combine the conformational search within the pattern identification process. In this approach the pattern identification module will be able to request from the conformational search module conformations that contain certain features of the ligand in a specified spatial arrangement.
Energy
Minimization
Redundancy
Elimination
clustering, poling

12. Feature Extraction Pharmacophoric Features Atoms Topological groups
Functional
groups
N, C, O
This stage is used to extract the features from the input ligands, which are relevant for the pharmacophore pattern discovery. The structure description to be used as input to the pattern discovery algorithms should contain only these extracted features. When deciding which features should be extracted and represented, one needs to decide on which structure level similarities are sought. There are three main levels of resolution for defining the features:
Atom-Based: One of the simplest ways to define a feature is by the 3D position of an atom, associated with the atom type.
Topological-Based: In some methods the atoms are grouped into topological features like phenyl ring and carbonyl group
Function-Based: In other methods the atoms are grouped into chemical functional features that describe the kind of interactions important for ligand-receptor binding. The most common functional groups are:
1. Hydrogen bond acceptor
2. Hydrogen bond donor
3. Base (positively charged at physiological pH 7)
4. Acid (negatively charged at physiological pH 7)
5. Aromatic ring
6. Hydrophobic group
The difference between the topological representation to the functional representation is that the resolution of the functional features is lower. For example, a phenyl ring is only one specific type of aromatic ring. Several topological features may have the same chemical function and thus can be classified as the same functional feature.
Note that the functional features are not mutually exclusive. For example, a hydroxyl oxygen can be classified as both a Hydrogen-bond acceptor and donor. In addition, Hydrogen-bond acceptor can also be negatively charged.
Phenyl ring Carbonyl group
HB bond acceptor / donor Acid / Base
Aromatic Ring Hydrophobic Group
Resolution
High
Low

13. Functional Groups
The simplest way is to represent the functional groups by the positions of their centers. The center of a base, acid, hydrogen bond donor or hydrogen bond acceptor is usually defined as the position of an actual atom. But, for a hydrophobic region or aromatic ring, the center is defined as the centroid of the group.
For a more accurate representation, additional geometrical attributes are assigned. For example, Hydrogen-bond acceptors and donors can be represented by a vector. Vector representation is more accurate than a point representation since it imposes an additional constraint by requiring that both the ligand feature and its complementary feature on the receptor will be on a single vector. Sometimes, defining the direction of a hydrogen-bond is not straightforward. In such a case, the direction range is represented either by a sample set of vectors or by a point and a sphere-patch on which the receptor point should be located . Two other examples are the representation of a hydrophobic group by a sphere and the representation of an aromatic ring by a plane and its normal.

14. Structure Representation
The features are combined to form a representation of the whole structure
3 main approaches:
3D point set
Graph
Set of interpoint distances
For each ligand structure the selected features are combined to form a representation of the whole structure. There are three main approaches:
3D point set: In most of the methods, a ligand structure is represented as a set of labeled points in three-dimensional space, where each point is associated with a feature.
Graph: Another approach is to represent a ligand structure as a labeled graph with nodes representing the features and the edges representing the relations. This being so, it is possible to use graph-theoretical methods for the identification of pharmacophoric patterns. For example, a molecule can be represented as a graph where atoms are vertices and bonds are edges.
Set of interpoint distances: Another approach is to consider a ligand structure to be a set of labeled points, together with the associated interpoint distances. In the Crandell-Smith method the considered points are atoms and the representation of each conformation contains all interatomic distances. This representation is stored in an nxn array commonly called distance matrix, where n is the number of atoms. The interpoint distance representation is orientation-independent, in contrast to the 3D point set representation. In addition, its main advantage comparing to 3D point set representation is that the necessary computations are simpler to carry out in distance space than in coordinate space. Comparing to the graph representation, the interpoint distance representation is actually a form of a complete graph in which vertices are labeled with atom properties and edges are labeled with distance information.

15. Pattern Identification
Goal definition:
MCS (Maximal Common Substructure): find the largest set of 3D features that common to all of the input ligands
MCS drawbacks: assume that there is a single common pharmacophore
Relaxed MCS: relaxing the requirement that all ligands must possess all features
A pattern or a configuration is a set of relative locations in 3D space, each associated with a feature. A ligand is said to match a pattern if it possesses a set of features and a conformation such that the set of features can be superimposed with the corresponding
Given a set of active ligands, the most popular approach to define the pattern that we are searching is MCS (Maximal Common Substructure). The pattern that we are searching is the largest set of pharmacophore features embedded in 3D space that is common to all of the input ligands.
MCS Drawbacks: The MCS approach is based on the assumption that there is a single common pharmacophore that is responsible for the observed activity. This assumption is not always correct. For example, it is often the case that a ligand can be highly active despite lacking a feature relevant in the binding of other ligands. In such a situation the requirement that the same set of features present in the same spatial arrangement in all of the ligands can lead to poor MCSs, e.g. MCSs that contain only a pair of features.
Relaxed MCS: The drawbacks of the MCS approach can be overcome by relaxing the requirement that all input ligands must possess all the features.

16. Pattern Identification (Cont.)
Relaxed MCS approaches:
A small number of ligands may miss a feature of the pharmacophore.
A pharmacophore should have at least M features in common with each ligand.
Methods:
Graph methods (Clique detection)
Exhaustive search
Genetic Algorithms (GA)
Relaxed MCS: certain ligands may be permitted to miss a feature as long as the total number of ligands missing a feature is below a predefined threshold. For example, in HipHop the user defines how many ligands must map completely or partially to the pharmacophore.
An alternative relaxed-MCS approach is to identify the smallest 3D configuration of pharmacophoric features that has at least m features in common with each of the input ligands. This approach is used in the MPHIL method. It differs from the MCS approach in that the input ligands are not required to share exactly the same configuration of pharmacophoric features. The pharmacophore model is a 3D configuration of K features, with which all the N input ligands have at least m features in common. K has to be as small as possible. Therefore, the optimal case is in which all ligands share the same 3D configuration of m features (i.e. K = m). The worst possible case is in which none of the N ligands share the same feature (i.e. K = N x m).
Methods:
Clique Detection: As was explained, a ligand can be regarded as a graph in which both the nodes and the edges have labels, corresponding to the features (e.g. atom types) and the relations (e.g. inter-atomic distances) respectively. The graph representation enables applying established graph-theoretical methods to the identification of pharmacophoric patterns.
Exhaustive Search: The patterns are identified by an exhaustive search, starting with small sets of features and extending them until no larger common pattern exists. Specifically, the algorithm starts with all possible combinations of two-feature patterns. Once all two-feature patterns are exhausted, it attempts to generate three-feature patterns by searching for a new feature to add to the two-feature patterns. The process terminates when no larger common pharmacophore patterns are recognized.
Genetic Algorithm (GA): A general randomized optimization technique. It is useful for solving combinatorial problems with search spaces that are too large for exploration by deterministic search algorithms. A genetic algorithm simulates the process of natural selection by manipulating a population of data-structures called chromosomes. Each chromosome represents a potential solution to a considered problem. Starting from an initial population of chromosomes, each generation undergoes different changes via genetic operators like mutation and crossover, which simulate the evolution. Each time a new generation of the population is created according to the 'survival of the fittest' principle, which ensures that over time the population should move toward the optimum solution.

17. Scoring
Requirement: the higher the scoring, the less likely it is that the ligands satisfy the pharmacophore model by a chance.
The size of a pharmacophore model can sometimes be misleading as a score:
2 charge features > 4 hydrophobes
Scoring is more complicated for relaxed MCS
In this stage we score and rank the pharmacophore candidates, obtained by the previous stages.
The basic requirement from a scoring scheme is that the higher the scoring, the less likely it is that the ligands satisfy the pharmacophore model by a chance correlation.
The size of the pharmacophore candidates can sometimes be misleading as a score. For instance, a charge center is more rare than a hydrophobic one. Therefore, a pharmacophore candidate of two charge centers can be more significant than a candidate of four hydrophobes.
In methods that relax the MCS requirement, the scoring scheme is more complicated. Since not all ligands must possess all of the pharmacophoric features, a pharmacophore candidate composed of three features shared by all ligands may be more significant than a candidate of four features shared by all ligands except for one.

18. Pharmacophore Fingertprints

19. Definition Also termed pharmacophore key
1D descriptor that encodes pharmacophoric information
e.g. encodes all the potential n-point pharmacophores that can be present in some conformer of a molecule.
Molecules can be represented by a pharmacophore fingerprint (also termed pharmacophore key). This fingerprint encodes pharmacophoric information.

20. Pharmacophore Space: The concept behind defining a pharmacophore fingerprint is to define a finite set of pharmacophores, pharmacophore space, that can be enumerated. For this purpose, Chem-X considers only n-point pharmacophores, where n is usually 3 or 4. A 3-point pharmacophore is defined by a set of three pharmacophore features, referred to as centers, and the three inter-center distances between them. By default, seven center types that are probably the most important for ligand-receptor interactions are defined: Hydrogen-bond donor, hydrogen-bond acceptor, positively charged center, aromatic ring centroid, lipophilic (hydrophobic) center, acidic center and basic center. In a similar way, a 4-point pharmacophore is defined by four centers and six inter-center distances. Furthermore, the continuous range of the inter-center distances is partitioned into a specified number of bins. The pharmacophore space is then defined by all combinations of n-point pharmacophores, together with all distance ranges for each of the inter-center distances.
Pharmacophore Fingerprint Denition: After defining the pharmacophore space, the pharmacophore fingerprint is defined as a binary bit string that indicates the presence or absence of potential pharmacophores in a molecule. Specifically, each bit in the fingerprint corresponds to a potential pharmacophore, and the value of the bit is set to one, if the pharmacophore is present in a molecule, or zero if it is absent.

21. Evaluating Molecular Similarity
A pharmacophore key for a set of ligands:
Union key = the logical OR of the keys of the individual compounds in the set.
Molecular Similarity:
e.g. Tanimoto coefficient:
Tc = NAB/(NA+NB-NAB)
0 ≤ Tc ≤ 1
Pharmacophore fingerprints are usually generated for a set of compounds, instead of individual one. For each low-energy conformer of each compound, every possible combination of 3 or 4 features is calculated and used to set the corresponding bit in the fingerprint. The obtained fingerprint is termed the union key and is simply the logical OR of the keys of the individual compounds in the set.
Molecular Similarity: The pharmacophore fingerprints are binary bit strings. Thus, the similarity of two fingerprints can be evaluated by counting the number of set bits in common or calculating similarity coefficients. An example is the Tanimoto coefficient, which is defined as Tc = NAB/(NA+NB-NAB) where NA is the number of bits set on in the fingerprint of molecule A, NB is the number of bits set on in the fingerprint of molecule B and NAB is the number of bits set on in common in the fingerprints of A and B. The values of this coefficient range from 1 for identical molecules and 0 for molecules that have no common bits

22. Pharmacophore Diversity
library diversity = #bits set in the union key
Designing a diverse library
Goal: reduce the number of molecules without decreasing the diversity
Process: Diverse Subset Selection
Rejecting too rigid or too flexible molecules
Calculating pharmacophore fingerprints
Rejecting molecules with too few or too many pharmacophores.
Iterative selection process
Pharmacophore Profiling
Pharmacophore fingerprints can be used to assess the diversity of a library, which is a very crucial in lead generation process. Specifically, the diversity of a library can be expressed by the number of different n-point pharmacophores it exhibits i.e., by the number of bits set in the union key of the library.
Furthermore, pharmacophore fingerprints can also be used to design a diverse library. The goal of this design is to reduce the number of molecules without decreasing the diversity. By avoiding molecules which are very similar, there is a potential of finding leads more rapidly because a smaller number of molecules are tested. In Chem-X the technique used to design a diverse library is termed diverse subset selection and it works in two stages as follows:
In the first stage molecules from the given set can be rejected by the following criteria:
Flexibility: Molecules with a few rotatable bonds are considered to be too rigid and are ignored. The reason is that too many such molecules are needed to cover the pharmacophore space. Furthermore, molecules which are too flexible are also ignored because they may possess too many pharmacophores and thus they can rapidly set on all the bits in the pharmacophore fingerprint. In addition, highly flexible molecules increases the required computational time. To speed up the process, the exibility criteria should be applied prior to generating the pharmacophore fingerprints of the molecules (which includes conformational analysis).
Number of Pharmacophores: These criteria are applied after generating the pharmacophore fingerprints of the molecules that pass the flexibility test. According to these criteria, molecules that possess too few or too many pharmacophores are rejected for the same reasons as before.
In the second stage an iterative selection process on the remaining molecules is performed. This process works as follows: The first molecule in the set being tested is automatically accepted into the diverse subset and its pharmacophore fingerprint is constructed. Then, the pharmacophore fingerprint of the next molecule is generated and its overlap with the fingerprint of the first molecule is calculated. The second molecule will be accepted into the diverse subset only if it exhibits a significant number of new pharmacophores not expressed in the first molecule. If it is accepted, the union key of the two molecules is computed by a logical OR operation. The remaining molecules will be considered in a similar manner, i.e. each molecule is compared to the union key of the previously selected molecules. As one can see, this technique is order dependent and there is no single correct answer.
Pharmacophore Profiling: The above definition of pharmacophore fingerprint assumes only a single bit to represent the presence or absence of a pharmacophore. However, a set of ligands that have been discovered to interact with the same receptor might have many pharmacophores in common. Moreover, those that occur frequently are more likely to be the actual pharmacophores that are responsible for binding. In these cases, one might be interested in counting the number of times a pharmacophore occurs within a set of molecules. Therefore, the original definition of pharmacophore fingerprint was extended by including a frequency count for each pharmacophore

23. Direct Methods for Pharmacophore Identication
1. Receptor-Based Approach
2. Complex-Based Approach

24. Negative Image of the Active Site
The negative image of the active site is used to construct a pharmacophore model.
Analyzes the features of the receptor active site and the spatial relationships among them.

25. Receptor-Based Approach
Negative Image: Functional features like hydrogenbond
donors/acceptors and lipophilic groups are identied in the active site. Then,
complementary features are placed within the binding site in chemically reasonable
positions.
4. Queries Generation:
a. The interaction map often displays a large number of features
b. Neighboring features of the same type are grouped to the
same cluster.
c. The feature, which is the closest to the geometric center of the cluster
is selected to represent the cluster, whereas the rest of the features are omitted.
d. Multiple queries are constructed

26. Complex-Based Approach
Provides information regarding the protein-ligand contacts.
Important!
1. The active site can be flexible and can rearrange itself to accommodate different ligands.
2. Alternative pharmacophores may be possible
within a single binding site.
3. Receptor may have more than one active site.

27. Multiple Alignment for Pharmacophore Investigation.

28. Pharmacophore Applications
1. Pharmacophore Searching.
2. De Novo Design of Ligands.
3. Lead Optimization.
4. 3D-QSAR.

29. Pharmacophore Searching
Pharmacophore searching is a part of a more general problem of 3D structure searching.
The main aspects in which methods differ from each other:
1. Pharmacophore Query Definition
2. Coping with Conformational Flexibility
3. Pattern Identification

30. Pharmacophore Query Definition
Pharmacophoric Features
Atoms
Topological
groups
Functional
groups
N, C, O
The representation influences the number of hits
The advantage of functional:
More hits
Different scaffolds
Phenyl ring Carbonyl group
HB bond acceptor / donor Acid / Base
Aromatic Ring Hydrophobic Group
Resolution
Low
High

31. Coping with Conformational Flexibility
Some methods combine a multi-conformational database with flexibility investigation.

32. Pattern Identification
Graph representation.
Solving the sub-graph isomorphism problem by reduction to clique detection.
Upper and lower bounds on the inter-feature distance constraints.

33. Multi-Level Searching
Main concepts:
Filter out (ASAP) the compounds that have no chance to satisfy the pharmacophore constraints.
The filters are applied in an increasing order of complexity, such that the first are fast and simple while successive ones are more time-consuming, but are applied only to a small subset.

34. Multi-level Framework
1. Screening
Presence of required features/atoms.
Comparison of keys/fingerprints.
2. Pattern Identification
Substructure searching.
3. Conformational Fitting
Only for ‘on the fly’ methods.
systematic search, random search, distance geometry, genetic algorithm and directed tweak.
A key can be simple and encode only the features present in a given structure.
More complicated keys encode distances, angles and even substructures, like n-point
pharmacophores. However, note that in systems where only a single conformer is
stored for each database molecule, the key of the molecule should represent not only
the stored conformer, but all its possible conformers. For example a key that encodes
upper/lower distance bounds can be suitable in this case. Another example is a
pharmacophore fingerprint that encodes all the potential n-point pharmacophores
that can be exhibited by the molecule in some conformation. Such keys will remove
molecules that can not adopt a conformation consistent with the geometry of the
pharmacophore. However, their efficiency at this stage can be argued due to the
high degree of flexibility exhibited by the molecules.

35. Receptor-Based Searching
Exploits structural information, like shape and volume. Docking techniques must fulfill:
1. Correct ranking
2. High speed
Docking techniques that require more than a few minutes per ligand molecule
are considered to be too computationally expensive to be used for screening [125]. Another
important requirement is the correct ranking of the ligand molecules. The relative
score of the top docking solution must be higher for ligands that bind well to the target
than for those that do not.

36. Approaches to Receptor-Based Pharmacophore Searching
1. Pharmacophore-Based Prescreening Prescreens the database using the pharmacophore information and then docks the selected candidates.
2. Pharmacophore-Constrained Docking
Incorporates pharmacophore information into the docking process.

37. Define the target binding site points.
Match the distances.
Calculate the transformation matrix for the orientation.
Dock and score the molecule.

38. Pharmacophore-Constrained Docking: FlexX-Pharm
1. Ligand fragmentation
2. Select & Place a set of base fragments
3. Construct the ligand by linking the remaining fragments.
New: A set of predefined pharmacophore requirements must be fulfilled!

39. De Novo Design of Ligands
A pharmacophore model can be used in De novo design to construct novel ligands that satisfy the physicochemical constrains.

40. De Novo Design vs. Searching
Availability 3D
Searching
Information
Existing molecules
Lead Generation
De Novo
Design
Novel molecules

41. Use of Pharmacophores in Lead Optimization
Lead optimization is the process in which a biologically active compound is modified to fulfill all the properties that are required from a drug (e.g. physicochemical and ADME/Tox properties).
Pharmacophore can direct chemists to include/exclude specific chemical groups.

42. QSAR Quantitative Structure Activity
Relationships (QSAR) uses statistical correlation methods, to predict quantities such as:
binding affinity
the toxicity
the pharmacokinetic parameters.
analyses the correlation between the structural features and the biological activity in order to predict the activity level of new componds.

43. Use of Pharmacophore in 3D-QSAR
3D-QSAR analyses the correlation between the structural features and the biological activity.
Pharmacophore models can be used to generate good alignments suitable for QSAR analysis.

45. Summary
Plays a key role in CADD, especially in the absence of receptor structure.
Can suggest a diverse set of compounds with different scaffolds.
Important!
1. Necessary but insufficient condition.
2. Several binding modes and several pharmacophores within the same active site.
3. Several active sites.

46. Pharmacophore-Based Prescreening Approach
Use the receptor active site to derive a pharmacophore query.
Search the DB of candidate ligands.
Dock the ligands into the receptor active site and score.