Drug Discovery by Design

(QSAR)
Gareth Thomas

05/28/24 prof. aza 1

 2.4 Quantitative Structure-Activity
Relationships (QSAR)
 The success of the SAR approach to drug
design depends not only on the knowledge and
experience of the design team but also on a
great deal of luck.
 QSAR is an attempt to remove the element of
luck from drug design by establishing a
mathematical relationship in the form of an
equation between biological activity and
measurable physicochemical parameters that
represent properties such as lipophilicity,
shape and electron distribution, which have a
major influence on the drug’s activity.
05/28/24 prof. aza 2
 These parameters are what are
currently thought to be a satisfactory
measure of the effect of these
properties on the activity of a
compound.
 They are normally defined so that they
are in the form of numbers that are
derived from practical data believed to
be related to the property that the
parameter represents.

05/28/24 prof. aza 3

 Consequently, it is possible to
either measure or calculate these
parameters for a group of
compounds and relate their values
to the biological activity of these
compounds by means of
mathematical equations using
statistical methods such as
regression analysis (see Appendix
1).
05/28/24 prof. aza 4
 Large numbers of these parameters
have now been measured and recorded
in the literature.
 Consequently, it is often possible to
calculate the theoretical value of a
specific parameter for an, as yet,
unsynthesised compound and, using the
experimentally establishes equation
relating activity to that parameter,
predict the possible activity of this
unknown compound.
05/28/24 prof. aza 5
 Alternatively the medicinal chemist
could determine from the equation the
value of the parameter and hence the
structures, that would give optimum
activity.
 Predictions of these types allow the
medicinal chemist to make a more
informed choice as to what analogues to
prepare. This could considerably cut
down the cost of drug development.
05/28/24 prof. aza 6
 The properties that have been found to
influence a drug’s activity are quite
diverse, the major ones being
lipophilicity, electronic effects and
shape.
 The parameters commonly used to
represent these properties are
partition coefficients for lipophilicity .
Hammelt a constants for electronic
effects and Taft M, steric constants
for steric effects.
05/28/24 prof. aza 7
 Consequently, this text will be largely
restricted to a discussion of the use of
these constants.
 However, the other parameters
mentioned in this and other texts are
normally used in a similar fashion.

05/28/24 prof. aza 8

 QSAR investigations are normally
carried out on groups of related
compounds that vary only in the nature
of their substituents; very few have
been carried out using structurally
diverse compounds.

05/28/24 prof. aza 9

 In QSAR relationships the
activity is normally expressed
as l/C (a concentration term),
usually C the minimum
concentration required to cause
a defined biological response.

05/28/24 prof. aza 10

  This means that an increase in activity
will correspond to an increase in the
value of 1/C.
 Furthermore, in most cases, to make
the numbers easier to handle, activity is
usually defined as log(1/C).
 As a result, QSAR relationships usually
take the general form:

05/28/24 prof. aza 11

 2.4.1 The Partition Parameters

 Two parameters are commonly used to

relate distribution with biological
activity, namely, the partition
coefficient (P) and the lipophilicity
substituent constant (π).
 The former parameter refers to the
whole molecule whereas the latter is
related to substituent groups.
05/28/24 prof. aza 12
 2.4.1.1 Partition Coefficients (P)
 A drug has to pass through a number of
biological membranes in order to reach its
site of action.
 Consequently, partition coefficients were the
obvious parameter to use as a measure of the
movement of the drug through these
membranes.
 The nature of the relationship obtained
depends on the range of P values for the
compounds used, If this range is small the
results may, by the use of regression analysis
(see Appendix 1), be expressed as a straight
line equation having the general form:
05/28/24 prof. aza 13
 where k1 and k2 are constants,
This equation indicates a linear
relationship between the activity of
the drug and its partition coefficient .
A number of examples of this type of
correlation are known (Table 2.4).

05/28/24 prof. aza 14

 Over larger ranges of P values the graph
of log(1/C) against log P often has a
parabolic form (Figure 2.10) with a
maximum value (log P°).
 The existence of this maximum value
implies that there is an optimum balance
between aqueous and lipide solubility for
maximum biological activity.
05/28/24 prof. aza 15
 Table 2.4. Examples of linear relationships between log(
1/C) against log P. Equations (ii) and (iii) are adapted
from C. Hansch, DRUG DESiGN 1, ed. E. J. Ariens,
copyright 1971 by Academic Press,and C. Hansch. r =
the regression constant (see Appendix 1). n = the
number of compounds tested and s = the standard
deviation.

05/28/24 prof. aza 16

 Figure 2.10. A parabolic plot for log(1/C)
against log P.
05/28/24 prof. aza 17
 Below P° the drug will be reluctant to
enter the membrane, whereas above P°
the drug will be reluctant to leave the
membrane.
 LogP° represents the optimum partition
coefficient for biological activity.
 This means that analogues with
partition coefficients near this optimum
value are likely to be the most active
and worth further investigation.
05/28/24 prof. aza 18
 Hansch and collegues showed that many
of these parabolic relationships could be
represented reasonably accurately by
equations of the form:

05/28/24 prof. aza 19

 The values of the constants in Equation
(2.3) are normally determined by either
regression analysis (see Appendix 1) or
other statistical methods.
For example, a study of the inducement
of hypnosis in mice by a series of
barbiturates showed that the
correlation could be expressed the
equation:

05/28/24 prof. aza 20

 This equation has a maximum log P at
about 2.0.
 Hansch and collegues showed that a
range of non-specific hypnotic drugs
with widely different types of
structure were found to have log P
values around 2.
 This implies that it is the solubility of
these different drugs in the membrane
rather than their structures that is the
major factor in controlling their
activity.
05/28/24 prof. aza 21
 On the basis of these and other
partition studies. Hansch suggested in
the mid-1960s that any organic
compound with a P value of
approximately 2 would, provided that it
was not rapidly metabolised or
eliminated, have some hypnotic
properties and be rapidly transported
into the CNS.
05/28/24 prof. aza 22
 Subsequent practical evidence gives
some support to this assertion.
 The fact that the thiobarbiturates have
log P values of about 3.1 suggests that
these drugs probably have a different
site of action from those of the
barbiturates. The larger value also
suggests that a more lipophilic receptor
is involved.

05/28/24 prof. aza 23

 The accuracy of the correlation of drug
activity with partition coefficient will
also depend on the solvent system used
as a model to measure the partition
coefficient values. The octanol/water
system is frequently chosen because it
has the most extensive database.
 However, more accurate results may be
obtained if the organic phase is
matched to the area of biological
activity being studied.
05/28/24 prof. aza 24
 For example, n-octanol usually gives the
most consistent results for drugs
absorbed in the gastrointestinal (GI)
tract, whereas less polar solvents such
as olive oil frequently give more
consistent correlations for drugs
crossing the blood—brain barrier.

05/28/24 prof. aza 25

 More polar solvents such as chloroform
give more consistent values for buccal
absorption (soft tissues in the mouth).
However, when correlating P values with
potency of activity it should be borne in
mind that the partition coefficient of a
drug is usually only one of a number of
parameters influencing its activity.

05/28/24 prof. aza 26

 Consequently, in cases where there
is a poor correlation between the
partition coefficient and the drug’s
activity, other parameters must he
playing a more important part in the
action of the drug.

05/28/24 prof. aza 27

2.4.1.2 Lipophilic Substituent Constants
(π)

05/28/24 prof. aza 28

 The values of π will vary depending on the
solvent system used to determine the
partition coefficients.
 However, most π values are determined using
the n-octanol/water system.
 A positive π value indicates that a substituent
has a higher lipophilicity than hydrogen and so
will probably increase the concentration of
the compound in the n-octanol layer and by
inference its concentration in the lipid
material of biological systems.
05/28/24 prof. aza 29
 Conversely, a negative π value shows
that the substituent has a lower
lipophilicity than hydrogen and so
probably increases the concentration of
the compound in the aqueous media of
biological systems.

05/28/24 prof. aza 30

 Table 2.5. Examples of the variations of π values
with chemical structure.

05/28/24 prof. aza 31

 The π value for a specific substituent will
vary with its structural environment (Table
2.5).
 Consequently, average values or the values
relevant to the type of structures being
investigated may be used in determining
activity relationships.
 Where several substituents are present the
value of π for the compound is the sum of the
π values of each of the separate substituents.

05/28/24 prof. aza 32

 For example, the π value for the two
methyl groups in 1,3-dimethylbenzene is:

05/28/24 prof. aza 33

 Lipophilic substituent constants can be
used as an alternative to the partition
coefficient when dealing with a series
of analogues in which only the
substituents are different.
 This usage is based on the assumption
that the lipophilic effect of the
unchanged part of the structure is
similar for each of the analogues
05/28/24 prof. aza 34
 Consequently, the π values of substituents
indicate the significance of the contribution
of that substituent to the lipophilicity of the
molecule.
 Furthermore, biological activity — π
relationships that have high regression
constants and low standard deviations
demonstrate that the substituents are
important in determining the lipophilic
character of the drug.
05/28/24 prof. aza 35
 Lipophilic substituent constants can also be
used to calculate theoretical partition
coefficients for whole molecules.
 These calculated values are often in good
agreement with the experimentally
determined values provided that the
substituents are not sterically crowded.
 For example, the calculated value of the n
octanol/water partition coefficient

05/28/24 prof. aza 36

 For 1.3-dimethylhenzene is 3.25, which is in
good agreement with the experimentally
determined value of 3.20.
 However, poorer agreements are usually found
when substituents are located close to each
other in the molecule.
 In addition, strong electron interactions
between substituent groups can result in
inaccurate theoretical P values.

05/28/24 prof. aza 37

 2.4.2 Electronic Parameters
 The distribution of the electrons in a
drug molecule will have a considerable
influence on the distribution and
activity of a drug. In order to reach its
target a drug normally has to pass
through a number of biological
membranes.
 As a general rule, non-polar and polar
drugs in their unionised form are usually
more readily transported through
membranes than polar drugs and drugs
in their ionised forms.
05/28/24 prof. aza 38
 Furthermore, once the drug reaches its
target site the distribution of electrons
in its structure will control the type of
bonds it forms with that target, which
in turn affects its biological activity.

05/28/24 prof. aza 39

 The first attempt to quantify the electronic
effects of groups on the physicochemical
properties of compounds was made by
Hammett in 1940.
 He studied the effect of substituents on the
position of equilibria of benzoic acid
derivatives. His proposals are now used
extensively as a measure of the electronic
effects of the structure of a drug on its
activity.

05/28/24 prof. aza 40

 2.4.2.1 The Hammett Constant (a)
 The distribution of electrons within a
molecule depends on the nature of the
electron- withdrawing and electron-
donating groups found in that structure.
For example, benzoic acid is weakly
ionised in water:

05/28/24 prof. aza 41

 Substitution of a ring hydrogen by an
electron-withdrawing substituent (X),
such as a nitro group, will weaken the 0-
H bond of the carboxyl group and
stabilise the carboxylate anion (Figure
2.11).
 This will move the equilibrium to the
right, which means that the substituted
compound is stronger than benzoic acid
(Kx > K).
05/28/24 prof. aza 42
 Conversely, the introduction of an
electron donor substituent (X), such as
a methyl group, into the ring
strengthens the acidic 0-H group and
reduces the stability of the carboxylate
anion.
 This moves the equilibrium to the left
which means that the compound is a
weaker acid than benzoic acid (Kx < Ks).
05/28/24 prof. aza 43
  Hammett used these concepts to calculate
substituent constants (σx) for a variety of ring
substituents (X) of benzoic acid using this acid
as the reference standard (Table 2.6).
 He used these constants, which are now known
as Hammett substituent constants or simply
Hammett constants, to describe the
relationship between linear free energy and
structure. Hammett constants were defined as:

05/28/24 prof. aza 44

05/28/24 prof. aza 45
 Figure 2.11. The effect of electron-
withdrawing and donor groups on the position
of equilibrium of substituted benzoic acids.

05/28/24 prof. aza 46

Table 2.6. Examples of the different
electronic substitution constants used
in QSAR studies.
Inductive substituent constants
(σ1) are the contribution that the
inductive effect makes to
Hammett constants and can be
used for aliphatic compounds.

05/28/24 prof. aza 47

 Taft substitution constants (σ*)
refer to aliphatic substituents,
but use propanoic acid (the 2-
methyl derivative of ethanoic
acid) as the reference point.

05/28/24 prof. aza 48

 The Swain—Lupton constants represent
the contributions due to the inductive
(F) and mesomeric or resonance (R)
components of Hammett constants.
 Adapted from An Introduction to the Principles of
Drug Design and Action by Smith and Williams, 3rd
Edd, ed. H. J. Smith, 1998. Reproduced by permission
of Harwood Academic Publishers.

05/28/24 prof. aza 49

 Different electronic substitution
constants used in QSAR studies.

05/28/24 prof. aza 50

05/28/24 prof. aza 51
 A negative value for σx indicates that
the substituent is acting as an electron
donor group because Kx > K.
 Conversely, a positive value for σx. shows
that the substituent is acting as an
electron-withdrawing group because K >
Kx
 Its value varies with the position of the
substituent in the molecule.

05/28/24 prof. aza 52

 Consequently, this position is usually
indicated by the use of the subscripts
o, m and p.
 Where a substituent has opposite signs
depending on its position on the ring it
means that in one case it is acting as an
electron donor and in the other as an
electron-withdrawing group.
05/28/24 prof. aza 53
 This is possible because the Hammett
constant includes both the inductive and
mesomeric (resonance) contributions to
the electron distribution.
 For example, the σm Hammett constant
for the methoxy group of m-
methoxybenzoic acid is 0.12, whereas
for p-methoxybenzoic acid it is —0.27.

05/28/24 prof. aza 54

 In the former case the electronic
distribution is dominated by the
inductive (I or F) contribution whereas
in the latter case it is controlled by the
mesomeric (M) or resonance (R) effect.

05/28/24 prof. aza 55

05/28/24 prof. aza 56
 Hammett postulated that the σ values
calculated for the ring substituents of a
series of benzoic acids could also be
valid for those ring substituents in a
different series of similar aromatic
compounds.
 This relationship has been found to be
in good agreement for the meta and
para substituents of a wide variety of
aromatic compounds but not for their
ortho substituents.
05/28/24 prof. aza 57
 The latter is believed to be due to
steric hindrance and other effects,
such as intramolecular hydrogen
bonding, playing a significant part in the
ionisations of compounds with ortho
substituents.
 Hammett substitution constants also
suffer from the disadvantage that they
only apply to substituents directly
attached to a benzene ring.
05/28/24 prof. aza 58
 Consequently, a number of other
electronic constants (Table 2.6) have
been introduced and used in QSAR
studies in a similar manner to the
Hammett constants.
 However, Hammett substitution
constants are probably still the most
widely used electronic constants for
QSAR studies.
05/28/24 prof. aza 59
 Attempts to relate biological activity to
the values of Hammett substitution and
similar constants have been largely
unsuccessful because electron
distribution is not the only factor
involved.
 One of the successful attempts to
relate biological activity to structure
using Hammett constants was the
investigation by Fukata and Metcalf into
the effectiveness of diethyl aryl
phosphates for killing fruit flies.
05/28/24 prof. aza 60
 This investigation showed that the
activity of these compounds is
dependent only on electron distribution
factors.
 Their results may be expressed by the
relationship:

05/28/24 prof. aza 61

  This equation shows that the
greater the positive value for σ, the
greater the biological activity of
the analogue.
 This type of knowledge enables one
to predict the activities of
analogues and synthesise the most
promising rather than spend a
considerable amount of time
synthesising and testing all the
possible analogues.
05/28/24 prof. aza 62
 2.4.3 Steric Parameters
 In order for a drug to bind effectively
to its target site the dimensions of the
pharmacophore of the drug must be
complementary to those of the target
site .
 The Taft steric parameter (Es) was the
first attempt to show the relationship
between a measurable parameter
related to the shape and size (bulk) of a
drug and the dimensions of the target
site and a drug’s activity.
05/28/24 prof. aza 63
  This has been followed by Charton’s
steric parameter (v),Verloop’s steric
parameters and the molar refractivity
(MR) amongst others. The most used of
these additional parameters is probably
the molar refractivity.
 However, in all cases the required
parameter is calculated for a set of
related analogues and correlated with
their activity using a suitable statistical
method such as regression analysis (see
Appendix 1).
05/28/24 prof. aza 64
 The results of individual investigations
have shown varying degrees of success
in relating the biological activity to the
parameter.
 This is probably because little is known
about the liner details of the three-
dimensional structures of the target
sites.
05/28/24 prof. aza 65
 Steric Parameter

Charton’s steric parameter (v)

05/28/24 prof. aza 66

 2.4.3.1 The Taft Steric Parameter (Es)

 Taft in 1956 used the relative rate constants

of the acid-catalysed hydrolysis of α-
substituted methyl ethanoates to define his
steric parameter because it had been shown
that the rates of these hydrolyses were
almost entirely dependent on steric factors.
He used methyl ethanoate as his standard and
defined Es as:

05/28/24 prof. aza 67

 where k is the rate constant of the
appropriate hydrolysis and Es = 0 when
X = H.
 It is assumed that the values for Es
(Table 2.7) obtained for a group using
the hydrolysis data are applicable to
other structures containing that group.
The methyl-based Es values can be
converted to H-based values by adding
1.24 to the corresponding methyl-based
values.
05/28/24 prof. aza 68
 Taft steric parameters have been found
to he useful in a number of
investigations.
 For example, regression analysis
Appendix 1 has shown that the
antihistamine effects of a number of
related analogues of diphenhydramine
(Figure 2.12) are related to their
biological response (BR) by

05/28/24 prof. aza 69

 Table 2.7. Examples of the Taft steric parameter Es

05/28/24 prof. aza 70

 Figure 2.12. The general formula of
Diphenhydramine analogues.

05/28/24 prof. aza 71

05/28/24 prof. aza 72
 A comparison of the standard deviations
(s) for Equations (2.12) and (2.13) shows
that the calculated values for the
Hammett constants a for each of the
analogues are more scattered than the
calculated values for the corresponding
Taft Es values.
 Furthermore, although both the r and s
values for Equation (2.12) are
reasonable, those for Equation (2.13)
are unacceptable.
05/28/24 prof. aza 73
 This indicates that the antihistamine
activity of these analogues appears to
depend more on steric than electronic
effects.
 This deduction is supported by the fact
that using regression analysis to obtain
a relationship involving both the
Hammett and Taft constants does not
lead to a significant increase in the r
and s values (Equation 2.14).

05/28/24 prof. aza 74

  Taft constants suffer from the
disadvantage that they are determined
by experiment.
 Consequently, the difficulties in
obtaining the necessary experimental
data have significantly limited the
number of values recorded in the
literature.

05/28/24 prof. aza 75

 2.4.3.2 Molar Refractivity (MR)

where n is the refractive index, M is

the relative molecular mass and ρ is the
density of the compound.

05/28/24 prof. aza 76

 The M/ρ term is a measure of the molar
volume whereas the refractive index term is a
measure of the polarisabilily of the compound.
Although MR is calculated for the whole
molecule, it is an additive parameter and so
the MR values for or a molecule can be
calculated by adding together the MR values
for its component parts (Table 2.8).

05/28/24 prof. aza 77

Tahe 2.8. Examples of calculated MR values.
Reproduced by permission of John Wiley and
Sons Lid. loin C. Hansch and A.J. Leo.
Substituents Constants for Correlation Analysis
in Chemistry and Biology. 1979.

05/28/24 prof. aza 78

 2.4.3.3 The Other Parameters

 These can he broadly divided into those that

apply to sections of the molecule and those
that involve the whole molecule.
 The former include parameters such as van
der Waals radii, Charton’s steric constants
and the Verloop steric parameters.
 The latter range from relative molecular mass
(RMM) and molar volumes to surface area.
They have all been used to corelate biological
activity to structure with varying degrees of
success.

05/28/24 prof. aza 79

 2.4.4 Hansch Analysis
 Hansch analysis is based on the attempts by
earlier workers, notably, Richardson (1867).
Richet(1893), Meyer (1899), Overton (1901),
Ferguson (1939) and Collander (1954), to
relate drug activity to measurable chemical
properties.
 Hansch and co-workers in the early 1960s
proposed a multiparameter approach to the
problem based on the lipophilicity of the drug
and the electronic and steric influences of
groups found in its structure.

05/28/24 prof. aza 80

 They realised that the biological
activity of a compound is a function of
its ability to reach and bind to its
target site.
 Hansch proposed that drug action could
be divided into two stages:
 (i) the transport of the drug to its site
of action
 (ii) the binding of the drug to the
target site.
05/28/24 prof. aza 81
  He stated that the transport of the
drug is like a random walk from the
point of administration to its site of
action. During this walk the drug has to
pass through numerous membranes and
so the ability of the drug to reach its
target is dependent on its lipophilicitv.
 Consequently,this ability could be
expressed mathematically as a function
of either the drug’s partition
coefficient or the π value(s) of
appropriate substituents.
05/28/24 prof. aza 82
 However, on reaching its target the
binding of the drug to the target site
depends on the shape, electron
distribution and polarisability of the
groups involved in the binding .
 A variety of parameters are now used to
describe each of these aspects of drug
activity, the most common ones being
the Hammett electronic σ and Taft Es
constants .
05/28/24 prof. aza 83
  Hansch postulated that the biological
activity of a drug could be related to all
or some of these factors by simple
mathematical relationships based on the
general format:

Log 1/C = k1 + k2 + k3 + k4

05/28/24 prof. aza 84

  where C is the minimum concentration
required to cause a specific biological
response, and k1, k2, k3 and k4 are
numerical constants obtained by feeding
the data into a suitable computer
statistical package.
 For example, the general equation relating
activity to all these parameters often
takes the general form:

05/28/24 prof. aza 85

  where other parameters could be
substituted for P, σ and Es. For
example, π may be used instead of P and
MR for Es.
 However, it is emphasised that these
equations, which are collectively known
as Hansch equations, do not always
contain the main three types of
parameter (Table 2.9).

05/28/24 prof. aza 86

 The numerical values of the constants in
these equations are obtained by feeding
the values of the parameters into a
suitab!e computer package.
 These values are obtained either from
the literature (eg π, σ, Es) or
determined by experiment (eg C, P etc).

05/28/24 prof. aza 87

 Table 2.9. Examples of simple Hansch equations.

05/28/24 prof. aza 88

 Many QSAR investigations involve varying
more than one ring substituent. In these
cases the values of the same parameter for
each substituent are expressed in the Hansch
equation either the sum of the individual
parameters or as independent individual
parameters.
 For example. in the hypothetical case of a
benzene ring with two substituents X and Y
the Hammet constants could be expressed in
the Hansch equation as either:
05/28/24 prof. aza 89
 k1∑(σx + σy) or k1σx + k2σy)
 comprehensive list of many of the
parameters used in Hansch analysis
may he found in a review by Tute
(Advances in Drug Research, 6. 1
(1971).

05/28/24 prof. aza 90

 Hansch equations can be used to give
information about the nature of the
mechanism by which drugs act, as well
as predict the activity of as yet
unsynthesised analogues.
 In the former case the value of the
constant for a parameter gives an
indication of the importance of the
influence that parameter in the
mechanism by which the drug acts.

05/28/24 prof. aza 91

  Consider, for example, a series of
analogues whose activity is related to
the parameters π and σ by the
hypothetical Hansch equation

05/28/24 prof. aza 92

 The small value of the coefficient for σ
relative to that of π in Equation (2.18)
shows that the electronic factor is not
an important feature of the action of
these drugs.
 Furthermore,the value of the
regression coefficient (r) for the
equation will indicate whether suitable
parameters were used for its
derivation.
05/28/24 prof. aza 93
 Values of r that are significantly lower
than 0.9 indicate that either the
parameter(s) used to derive the
equation was unsuitable or that there is
no relationship between the compounds
used and their activity.
 This suggests that the mechanisms by
which these compounds act may be very
different and therefore unrelated.
05/28/24 prof. aza 94
 Predictions of the activities of as yet
unsynthesised analogues are useful in
that they allow the medicinal chemist to
make a more informed choice as to
which analogues to synthesise

05/28/24 prof. aza 95

 However, these predictions should only
be made within the limits used to
establish that relationship.
 For example, if a range of partition
coefficients with values from 3 to 8
were used to obtain an activity
relationship — partition coefficient
equation then this equation should not
be used to predict activities of
compounds with partition coefficients
of less than 3 and greater than 8.
05/28/24 prof. aza 96
 Hansch equation activity
predictions that are widely
different from the observed values
suggest that the activity of a
compound is affected by factors
that have not been included in the
derivation of the Hansch equation.

05/28/24 prof. aza 97

 The discovery of this type of anomaly
may give some insight into the
mechanism by which the compound acts.
For example, a study by Hansch on the
activity of penicillins against a strain of
Staphylococcus aureus in mice gave the
in vivo relationship

05/28/24 prof. aza 98

 This relationship (2.19) predicts that a
penicillin with branched side chain
PhOCH(CH3)CO— should be less active
than a penicillin with the similar sized
unbranched side chain

05/28/24 prof. aza 99

 Figure 2.13. Penicillin analogues.

05/28/24 prof. aza 100

 This relationship (2.19) predicts
that a penicillin with branched side
chain PhOCH(CH.)CO— should be
less active than a penicillin with the
similar sized unbranched side chain
PhOCH2CO— (Figure 2.13) because
the branched side chain has a
higher π value. However,

05/28/24 prof. aza 101

 However, Hansch found that the
penicillin with the branched side chain
was the more active.
 He suggested that this anomaly could be
explained by the fact that branched
side chains were more resistant to
metabolism than straight side chains.

05/28/24 prof. aza 102

 The accuracy of Hansch equations and hence
the success of OSAR investigations depend on
using enough analogues, the accuracy of the
data and the choice of parameters:
 (i) The greater the number of analogues used
in a study, the greater the probability of
deriving an accurate Hansch equation. A rough
rule of thumb is that the minimum number of
compounds used in the study should be not
less than 5x, where x is the number of
parameters used to obtain the relationship.

05/28/24 prof. aza 103

 Where substituents are being varied it
is also necessary to use as wide a
variety of substituens is in as large a
range of different positions possible.

05/28/24 prof. aza 104

 (ii) The accuracy of the biological data
used to establish a Hansch equation will
affect the accuracy of the derived
relationship because its value depends
to some extent on the subject used for
the measurement.
 Consequently, it is necessary to take a
stastically viable number of
measurements of any biological data
such as activity and use an average value
in the derivation of the Hansch
equation.
05/28/24 prof. aza 105
 Furthermore, extreme parameter values
should not be used because they are
likely to dominate the regression
analysis and so give less accurate
Hansch equations.
 Their presence suggests that the
parameter is either not suitable or
there is no possible correlation.
05/28/24 prof. aza 106
 (iii) The choice of parameter is
important because one parameter may
give an equation with an acceptable
regression constant whereas another
would give an equation with an
unacceptable regression constant.
 For example, it may be possible to
correlate successfully the dipole
moments of a series of analogues with
biological activity but fail to obtain a
correlation for the same series of
analogues using Hammett constants as a
parameter.
05/28/24 prof. aza 107
 Furthermore, some parameters like π
and o are interrelated and so their use
in the same equation can lead to
confusion in interpretating the equation.

05/28/24 prof. aza 108

 2.4.4.1 Craig Plots
 Craig plots are two-dimensional plots of one
parameter against another (Figure 2.14).
 The plot is divided into four sections
corresponding to the positive and negative
values of the pameters.
 They are used, in conjunction with an already
established Hansch equation tor a series of
related aromatic compounds, to select the
best aromatic substituents that are likely to
produce new highly active analogues.

05/28/24 prof. aza 109

 For example, suppose that a Hansch
analysis carried out on a series of
aromatic compounds yields the Hansch
equation:

05/28/24 prof. aza 110

Figure 2.14. An example of a Craig plot of para
Hammett constants agaist para π values.
Reprinted by permission of John Wiley and Sons.
Inc. from P. N. Craig. in Burgers Medicinal
Chemisry, ed. M. E. WolfE. 4th Ed. Part I, p.343.
1980. Copyright © John Wiley & Sons Inc.
05/28/24 prof. aza 111
05/28/24 prof. aza 112
 To obtain a high value for the activity
(I/C) and, as a result, a low value for C.
it is necessary to pick substituents with
a positive π value and a negative σ value.
In other words, if high- activity
analogues are required, the substituents
should be chosen from the lower right-
hand quadrant of the plot.

05/28/24 prof. aza 113

 However, it is emphasised that the use
of a Craig plot (does not guarantee that
the resultant analogues will be more
active than the lead because the
parameters used may not be relevant to
the mechanism by which the analogue
acts.

05/28/24 prof. aza 114

 2.4.5 The Topliss Decision Tree
 The Topliss decision tree is essentially a flow
diagram that in a series of steps direct the
medicinal chemist to produce a series of
analogues, some of which should have a
greater activity than the lead used to start
the tree. It is emphasised that only some of
the compounds will be more active than the
lead compound.

05/28/24 prof. aza 115

 The method is most useful when it is
not possible to make the large number
of compounds necessary to produce an
accurate Hansch equation.
 It effectively reduces the number of
compounds that a medicinal chemist
needs to synthesise in order to discover
potent analogues.

05/28/24 prof. aza 116

 Two decision trees are in use, one for
substituents directly attached to an
aromatic ring and the other for
aliphatic side-chain substituents of an
aromatic ring system (Figure 2.15).

05/28/24 prof. aza 117

 a
05/28/24 prof. aza 118
05/28/24 prof. aza 119
 Figure 2.15. The Topliss decision trees for (a)
an unfused aromatic ring and (b) an aliphatic
side chain (L = significantly lower activity. E =
about the same activity and M = significantly
greater activity). Reprinted with permission
from J. G.Topliss, Utilization of operational
schemes for analog synthesis in drug design.
Journal of Medicinal Chemistry, 15 (10), 1006
(1972). Copyright 1972 American Chemical
Society.
05/28/24 prof. aza 120
 They were constructed using
parameters similar to those described
earlier in this chapter and are both
used in a similar manner. However, the
use of the aromatic substituent decision
tree requires:
 (i) the lead compound to have an
unfused aromatic ring;
 (ii) the biological activity of the lead
compound and its analogues to be
readily determinable.
05/28/24 prof. aza 121
 To use the aromatic decision tree the
4-chloro analogue of the lead is
synthesised and its activity compared
with that of its lead.
 This activity may be significantly less
(L), approximately the same (E) or
significantly greater (M) than that of
the original lead.
 If the activity is greater than that of
the lead the next analogue to be
prepared is the next one on the M
route, namely the 3,4-dichloro analogue.
05/28/24 prof. aza 122
 Alternatively, if the activity of the
analogue is less than that of the original
lead the next step is to produce the 4-
methoxy analogue as indicated by the L
route on the tree.
 Similarly, if the activity is about the
same as that of the original lead the E
route is followed and the 4-methyl
analogue synthesised.

05/28/24 prof. aza 123

 This procedure is repeated, the activity
of each new analogue being compared
with that of the preceding compound in
order to determine which branch of the
tree gives the next analogue

05/28/24 prof. aza 124

 Suppose, for example, that a compound A
(Figure 2.16) is active against S. aureus and
activity of this compound and its analogues
can be readily assessed by a biological
method.
 The first step in the Topliss approach is to
synthesise the 4-chloro derivative (B) of A.
Suppose that the activity of B is greater than
that of A, then following the M branch of the
Topliss tree (Figure 2.16) indicates that the
next analogue to produce is the 3,4-dichloro
derivative (C) of A.
05/28/24 prof. aza 125
Figure 2.16. A hypothetical example of the use of the
Topliss decision tree. The compounds are srnthesised in
the order A. B, C... etc. It should be realised that only
some of the compounds synthesised will he more potent
than
05/28/24 the original lead A. prof. aza 126
 Once again suppose that the biological
assay of C was less than that of B.
 In this case, the Topliss tree shows
that the next most promising analogue
is the 4-trifluromethyl derivative of (D)
of A. At this point one would also
synthesise and biologically test the 2,4-
dichloro (F) and the 4-nitro analogues
(F) of A.

05/28/24 prof. aza 127

 It is emphasised that the decision tree is not
a synthetic pathway for the production of
each of the analogues.
 It simply suggests which of the substituents
would be likely to yield a more potent
analogue.
 The synthetic route for producing each of the
suggested analogues would vary for each
analogue and would use the most appropriate
starting materials.
05/28/24 prof. aza 128
 The Topliss decision tree does not give
all the possible analogues hut it is likely
that a number of the most active
analogues will be found by this method.
 Consider, for example. the anti-
inflammatory aryltetrazolylalkanoic
acids (Figure 2.17).
 The traditional approach required the
synthesis of 28 analogues to discover
the four most active analogues.
05/28/24 prof. aza 129
05/28/24 prof. aza 130
  Figure 2.17. A hypothetical example of
the use of the Topliss decision tree to
find active aryltetrazolvlalkanoic acids.
The analogues found by a traditional
approach to have a higher activity and
potency than the lead are indicated by
an asterisk.

05/28/24 prof. aza 131

 However, the use of the Topliss decision
tree would have yielded three of the
most active components by the
synthesis of only eight analogues, a
considerable saving on time and money.

05/28/24 prof. aza 132

 A limitation of the Topliss method is the
requirement that the lead must contain an
unsaturated aromatic ring system. However,
because many of the drugs in current usage
fulfil this requirement.
 The Topliss decision tree method can be of
some considerable use in the discovery of
more effective new analogues.

05/28/24 prof. aza 133

05/28/24 prof. aza 134
05/28/24 prof. aza 135
05/28/24 prof. aza 136
05/28/24 prof. aza 137
 Topic One
 Details about this topic
 Supporting information and examples
 How it relates to your audience

05/28/24 prof. aza 138

 Topic Two
 Details about this topic
 Supporting information and examples
 How it relates to your audience

05/28/24 prof. aza 139

 Topic Three
 Details about this topic
 Supporting information and examples
 How it relates to your audience

05/28/24 prof. aza 140

 Real Life
 Give an example or real life anecdote
 Sympathize with the audience’s situation if
appropriate

05/28/24 prof. aza 141

 What This Means
 Add a strong statement that summarizes
how you feel or think about this topic
 Summarize key points you want your
audience to remember

05/28/24 prof. aza 142

 Next Steps
 Summarize any actions required of your
audience
 Summarize any follow up action items
required of you

05/28/24 prof. aza 143