IL Drug Development Lead Modification
`
`the concentration of the drug that produces some standard biological effect,
`was related to its lipophilicity by the parabolic expression shov.•n in Eq.
`(2.6). 41
`
`log 1/C = -k(log P)2 + k'(log P) + I<'
`(2,6)
`Ort the basis of Eq. (2.5), it is apparent that if a compound is m~re soluble in
`water than in l~octmml, P is less tllan 1, and, therefore, log P is negative.
`Conversely, a molecule more soluble in 1-octanol has a P value greater than l ,
`and log P is positive, The larger the value of P, the more there will be an
`interaction of the drug with the lipid phase (i.e., membranes). As P ap(cid:173)
`p~esinfinit:}l,- lhe drug interaction will become so great that th"C"'drug~wftt
`net be able to cross the aqueous phase. and it will localize in the first lipop!hilic
`phase with which it comes into contact. As.f...4.Wroaef1es-zero,. the drug will
`be so water soluble that it will not be capable of crossing the lipid phase and
`will localize in the aqueous phase. Somewhere between P = 0 and P = oo,
`there will be a value of P such that drugs having this value will be least
`hindered in their journey through macromolecules to their site of actjon. This
`value !s called log P9 , the optimum partition coefficient for biological activity.
`This random walk analysis supports the parabolic relationship [Eq. (2.6)1
`between potency (log l/C) and log P (Fig. 2 .4). Note the correlation of Fig. 2.4
`with the generalization regarding homologous series of compounds (Section
`U.D,1; Fig. 2.1), An in<.:rease in the alkyl chain length increases the lipoph.ilic(cid:173)
`ity of the molecule; apparently. the log P 0 generally occurs in the range of 5-9
`carbon atoms. Hansch el al.41 found that a number of series of nonspecific
`hypnotics had similar lpg P0 valu.es. ap. prox. imately 2, and they.suggested that
`this is the value of log P () .needed for penetration into the central nervous
`system {CNS}. If a hypnotic .agent has a log P considerably different from 2,
`then its activity probably is derived from mechanisms other than just lipid
`
`log P
`
`Figure 2.4, Effect of log P on blologicaJ response. P is the partition coefficient, an.d C ls the
`concentration of the compound required to produce a st.an-Oard biological effect_
`
`

`
`2. Drug Discovery, Oe~;ign, and Developmen1
`
`tnlnsporL If a lead compound has modest CNS activity and has a log P value
`of 0, it would be, rt•asonable to synthesize an analog with a higher log P.
`Can you predict what analog will have a higher lo~rP? In the same \Vay that
`!ltibstltueut constants were derived by Hammett for the electronic effects of
`atoms {lfld groups (er c.onstants), Hansch and co~workerst9,:n.n derived sub~
`stituent constants for the contributi<m of individual atoms :tnd groups to the
`partition coefficient. The lipophilidty substiment consumt, Tr, is de.fined hy
`Eq. {2.7), which has the same derivation as the Hammett equation. The term
`Px is the partition .coefikient for the compound \Vilh substituent. X, and Pn is
`the partition coefficient for the parent molecule (X = .H). As in the case of the
`Harmnett substituent constant o-t 1i is additive and constitutive. Additive
`means that multiple substituents e;xert an influence ei:1ual to the sum of the
`individual substituents. Conszitutive indicates that the effect of a substituent
`may differ depending on the moiecule to which H is attached or on its environ(cid:173)
`ment Alkyl groups are some of the- least constitutive. For example, methyl
`groups anached al ihe meta or para positions of 15 different benzene deriva(cid:173)
`tives had ·trcJ1i values with a mean and standard derivation of 0.50 -:::: 0.04.
`Because of the additive nature of 1T' values. trcth can be determined as shown
`in Eq, (2.8), where the log P values are obtained from standard tabk:s. 42
`Because, by definition, ITH = 0, then 1TCtt1 = 1TCH)'
`J Px
`..
`·p
`-rr = log Px - Jog H = og ._,--PH
`7i'CHz = log Pni!f'!;}f'.lhane - Jog P11ini:•111t1hane
`(--0.33) = Q.51
`·= CU8 -
`
`(2.8)
`
`(2.7)
`
`As was alluded to in Section ll,D,2 on molecular modification, branching
`i~X!£~!tJowers the log P or Tl' as .. <t res~It t>f the larger molarvoluiiies
`.a,J!s! SQ.!lP:C~ _QfJ1ranch~.d compounds. As a ml~ {)f thumb, the value of log p
`or n is lowered by 0.2 unit per branch. For example, the 1iH:r value in
`3~isop:ropylphenoxyacetic add is l .30; trvr is 3(0.5) = l .50. Another case
`when: tr values arc fairly constant is conjugated systems, as exemplified by
`7TCH=CHCH.,.=CH in Table 2.5.
`Inductive effects are quite important-to HpophUidty.43 ln general, eJ~ctron~
`~~itbdrawing groups im.:rease '11' w,hen ~.b¥!,irogen~bon(iing group is involved.
`For example ·rrctt,oH varies as a function of the proximity of an electron(cid:173)
`withdrawing phenyl group lEq. (2.9)].44 and 1TN<Jz varies as a fum:::tion of the
`indm::tjve effect of the nitro group on the hydroxyl group {Eq. (2.10)]. 43 The
`electron~withdrawing inductive effects oftbe phenyl group (Eq. (2.9)} and the
`nitm b"l'OUp (Eq. (2.10)) make the nonbonded eJectnms on the hydrox}1J group
`less available for hydrogen bonding, thereby reducing the affinity of this func(cid:173)
`tional group for the aqueous phase. This, then) increases the log P or 1f. Alsu
`
`l
`I~
`
`

`
`31
`
`![)
`
`H
`
`2.14-0.75 ""
`
`! 39
`
`l.o3-0.6S "'
`
`1.38
`
`4.U-2.61 =
`
`'
`!AS: i
`
`::ii U2-Ull
`
`L3l
`
`~.45-2.13 "' 132
`
`-- JcgP
`
`m- Jog p
`
`-
`
`log P
`
`··-·log&"
`
`-
`
`logP
`
`~
`
`logP or9 .
`logP (JL)
`log p C;V
`
`~
`~d
`![)
`0
`
`t=
`
`21> (4.13}
`
`! 42
`
`)~
`f
`
`log P
`
`-
`
`logP
`
`0y-OH
`
`v
`
`lJM ..... I A6 '"'
`
`.L1i:l
`
`note in Eqs. (2.9) and (2.10) that .. because n'M = 0 by definition, iog P~e =
`
`11'~.
`
`log PPhC.ttl = - I . :n
`1fcthoH :::::. log PFhccnahOH -
`1rctt~ou = log Pr.1icrr2mr -
`log PPhH = -t.03
`'ITNO~ = logP~nNO~ -logPPhH :::t -0.28
`1TNQz = log p 4-NOtPliCH~ ~ log p~H,OH = 0.11
`Resonance effects also are important lo the Hpophilicity much the same
`way as are inductive effects.43 Deloc..alizatfom of nonbondcd electrons into
`aromatic systems decreases their availability for hydrogen bonding with· the
`
`(2.10)
`
`(2.9)
`
`

`
`,.•r""'"'···~-•'•·'·""···
`
`··c .... 1·.,.,'.•"'""""'"'""""'"""
`
`wx (aliphatic )/J
`
`llwx
`
`Table 2.6 Eff~t of Folding of Alkyl ChaJns on nA3
`x
`"'If,. (aromatic)"
`OH
`F
`Cl
`Br
`l
`COOH
`C02CHJ
`COCH3
`NH2
`CN
`OCH3
`CONH2
`
`-I.80
`-D.73
`-·O.f3
`OJ.'!4
`0.22
`·-1.2()
`-0.91
`'""l.26
`-J.85
`·-L47
`~0.98
`-2.28
`
`-1.16
`-o.n
`iJ.39
`o.ro
`l.61';
`-0.67
`--0.27
`-·0;7!
`-l.!9
`·-0'.811
`-0.47
`-l.71
`Average
`
`0.64
`ilSO
`0.52
`0.56
`0.78
`iL59
`0.64
`0.55
`0.66
`0.63
`0.51
`0.57
`{L60::t 0.05
`
`a Log Pl"l;(Ci'hllx -
`"Log PcHJICH;);x -
`
`!O'J PPh(CH?JJH •
`log PcH;(CHinH·
`
`aquwus phase and. therefore. incri?ases the 11. This is supported by the gen(cid:173)
`e:ral trend that aromatic '1l'X values are greater than aHphatic nx values, again
`emphasizing the constitutive natun: of 'TT and Jog P.
`Sterle effects are variable. 43 If a group sterically shields nonbonded elec(cid:173)
`trons, then aqueous interactions will decrease, and the '11' value wm increase.
`However, crowding offunctional groups involved in hydrophobic internctions
`(see Chapter 3) will have the opp0site effecl. Conformational effects altm can
`affect the 1f value.43 The 7tx values for :Ph(CH1)lX are .consistently lower
`(more water sofuble) than wx values for CJ-h(Cfl2)JX (fable 2,6). This phe(cid:173)
`n-omenonJs believed to be the result of folding of the side chain onto the
`phenylri.ng (2.39). which means a smaller apolar suJface for organic solvation.
`The folding may be caused by the interaction of the CH2"!'X dipole with the
`phenyl w electrons and by intramolecular hydrophobic interactions.
`
`CH4
`X
`"-cH; \
`~CH2
`
`2.39
`
`Two examples follow to show the additivity Df 1'l constants in predicting
`Jog P values. A calculation of the log P for the anticancer drug diethylstilbes(cid:173)
`trol (2.40) is as foUows:
`Cale. Jog P ~ 21l'rn1 + 21TcH2 + ucH=CH + 2 Jog PMH - 0.40
`= 2(0.50) + 2(0.50) + 0.69 + 2(1.46) - 0.40
`-::;: 5.21
`
`(2.11)
`
`

`
`It Drug Development; Lead Modification
`
`33
`
`ln Eq. (2.11), 1l"cH=CH = i('1Tcn=cHcn"""cH)., which was_ shown in Table 2.5 to
`be l(l .38); -OAO is added into the equation to account for two branching
`points (each end of the alkene). The calculated log P value of 5_2l is quite
`remarkable considering that the experimental log P value is 5.07.
`A calculation of log P for t.he antihistamine diphenhydramine (2.41) is
`shown .in Eq. (2.12). fo Eq. (2.12), 2.13 is log P for benze11e, which is the same
`as 'lTph; 0.30 is vcr:d0.50) - 0.20 for branching; -0.73 was ()btained by sub~
`tracting L50 (27rCH3 + 'i'TCH;) from log Pcn,ctt1ocHaCH, (=0.77); and -0.95 is the
`value for '11NMei obtained from Ph(CH2)3NMe2 _43 The experimental log P value
`is 3.27.
`
`q
`
`,,c~
`
`.u-(CH-OCIIz-~--CH;i-N '·
`\.=]
`'CH:t
`
`;?.41
`Cale. log P """ 21Tr1i + 1Tctt + 11c0Hi + rrcH1 + 1'fNMe,
`= 2(2. 13) + 030 - 0.13 + 0.50 - 0.95
`= 3.38
`
`(2.12)
`
`The chore of calculating Jog P values for molecules has been lessened
`considerably by the computerization of the method. 45 A nonlinear regression
`model for the estimation of partition coetlidents was developed by Bodor el
`al.46 using the following molecular descriptors: molecular surface, volume,
`weight, and charge densities. It was shown to have excellent predictive power
`for the estimation of log P for complex molecules.
`Although the log P values determined from l-octanot/water partitioning are
`excellent models for in vivo lipophilicity. it has been found for a variety of
`aromatic compounds with log P values exceeding 5.5 (very lipophilic) or
`molar volumes greater than 230 cm3/mol that there is a breakdown in the
`correlation of these values with those determined from partitioning between
`L-a·phosphatidylcholine dimyristoyl membnme vesicles and water.47 Above a
`log P value of 5.5 the solvent solubility for these molecules is greater than
`their membrane solubility. As the compound increases in size more energy
`per unit volume is required to form a cavity in the structured membrane
`
`

`
`34
`
`2. Drug Discovery, Design, and Development
`
`phase. This is consistent \vith observations that branched molecules have
`lower log P values than their straight chain counterparts and that this effect is
`even greater in membranes thari in organic solvents.
`It should be noted that although log P values are most commonly deter·
`mined with 1-octanol/water mixtures, this is not universal. For example.
`Sei1e1...is introduced a m:\v additive constitutive substiruent constant for sol(cid:173)
`vents other than 1-octanol. Therefore, when using fog P values, it is important
`to be rnvare of the solvent used to obtain the log P data.
`
`c, Steric Effects: '.fhe Taft .Equatfon. Since interaction of a drug with a
`receptor involves the mutual approach of two molecules, another important
`parameter for QSAR is the steric effect. In much the same way that Hammett
`derived quantitative electronic effects (see Section U~E,2.a), Taft49 defined
`the steric parameter Es [Bq. (2.B)J. Taft used for the reference reaction the
`relative rates of the acid~catalyzed hydrolysis of a~substittited acetates
`{XCH4C02Me). this parameter is normally standardized to the methyl group
`(XCH:z = CH3) so that Es(CH3) = o~o; it is possible to standardize it to hydro(cid:173)
`gen by adding 1.24 to every methyl-based Es value.so Hancock et al.51 claimed
`that this model reaction was under the influence of hyperconjugative effects
`and, therefore. developed corrected E,, values for the hyperconjugation of
`a--hydrogen atoms [Eq. (2.14)], where Esc is the corrected Es value and n is the
`nurnber of a-hydrogen atoms.
`
`(2.13)
`
`(2.14)
`Two other steric parameters worth mentioning are molar refractivity (it:!!?)
`and the Verloop parameter. Molar refractivity52 is defined by the Lo.rentz(cid:173)
`Lorenz equation fBq. (2.15)], where n is the index of refraction at the sodium
`D line, MW .is the molecular weight, and d is the density of the compound.
`The greater the positive :Ar!R value of a substituent, the larger is its sterk or
`bulk effect. This parameter also measures the electronic effect and, therefore,
`may re:fiect dipole-dipole interactions at the receptor site.
`MR :::: n2 -' 1 ~!W
`{">.15)
`112 +1 d
`'
`.
`'The Ver/Oop steric: parameters53 are used fo a program called STERIMOL
`to calculate the steric substituent values from standard bond angles, van dcr
`Waals radii, bond lengths, and user~determined reasonable conformations.
`Five parameters are involved. One (L) is the length of the substituent along
`the axfa of the bond between the substituent and the parent molecule. Four
`width parameters (Bi-B4) are measured perpendicular to the bond axis. These
`
`

`
`H. Drug Development Lead Modification
`
`five parameters describe the positions! relative to the point of attachment and
`the bond axis, of five planes which closely surround the group. In contrast to
`Ei values which, because of the reaction on which they are based, cannot be
`detennined for many substituents, the \! erloop parameters are available for
`any substituent
`
`3. Methods Used to Correlate Physkocbcmical Parameters with
`Biofogical A1;tivity
`Now that we can obtain numerous physicochemical parameters (also called
`descriptors) for any substituent, how do we use these parameters to gain
`inforn1ation regarding what compound to synthesize next ln an attempt to
`optimize the lead compound? .Hrst, several (usually, many} compounds r&
`lated to the lead are synthesized, and the biological activities are determined
`in son1e bioassay, These data, then, can be manipulated by a number of QSAR
`methods. ·The most popular is Hansch analysis.
`
`a. llansch Analysis: A Linear Multiple Regressim~ .4nalysis. With the
`realization that there are (at least) two considerations for biologie.:al activity,,
`namely, lipophilicity (required for the journey of the~ drug to the site of action)
`and electronic factors (required for drug interaction with the site of action),
`and that .lipophilicity is a parabolic function, Hansch and FujitaJ7 expanded
`Eq. (2.6) to that shown in either Eq. (2.l6a) or (2.16b) kno\Vfl as the Ho.nsclt
`equation, where C is the molar concentration (or dose) that elicits a standard
`biological response (e.g., ·ED51.i, the dose required for 50% of the maximal
`effect, ICs01 the concentration that gives 50% inhibition of an enzyme or
`antagonism of a rn<::eptor; or LDso; the lethal dose for 50% of the animal
`population). The tenns k, k', p,. and k' are the regressi-0n coefficients derived
`from statisrical curve fitting, and 1r and o- are the lipophilicity and electronic
`substituent constants, respectively. The reciprocal of the concentration (l/C)
`reflects the fact that greater potency is associated with a lower dose, ;im:I the
`negative sign for the w'l [or (log P)2] term reflects the expectation of an
`optimum lipopbilicity, that is, the -;r0 or tog P0 •
`log J/C = -hr 1 + k'1T + p<Y + !('
`log vc = -k(Jog p)Z + k'(log P) + fJff + K'
`Because of th~ importance of steric effects and oth~r shape factors of
`molecules for receptor interactions, an E$ term and a variety of other shape~
`size. or topography terms (S) have been added to the Hansch equation [see
`Eq. (2.17)]. The way these parameters are used is by the application of the
`method of linear multiple regressfr:>n anatvsis .54 The best least squares fit of
`the dependent variable (the biological activity) to a linear combination of the
`independent variab.les (the descriptors) is determined. Hansch anaiy:i'is, also
`
`(2.16a)
`(2.16b)
`
`

`
`So
`
`2. Drug Discovt.try, Design, and Development
`
`(2. 17)
`
`called tbe extrathemwdynamic method, then, is a linear free energy approach
`to drug design in congeneric series in which equations are set up involving
`different combinations of the physicochemical parameters; the statistical
`methodology allows the best equation to be selected and the statistical signifi(cid:173)
`cance of the correlation to be assessed. Once this equation has been estab(cid:173)
`lished, it can be used to predict the activities of untested compounds. Prnb(cid:173)
`Jems associated with the use of multiple regression analysis in QSAR studies
`have been discussed by Deardon.55
`log 1/C = -an 2 + h1T + po-+ <·Es + dS + e
`Several assumptions must be made when the extrathermodynamic method
`is t1lilizcd: conformational changes in receptors can be ignored, metabolism
`does not inte1fere, linear free energy terms reftwant to receptor affinity are
`add.itive, the potency-Hpophilicity relationship is parabolic or1inear, and cor(cid:173)
`relation implies a causal relatii1nship. According to Martin56a and Tute560 there
`is a balance of assets and Habilities to the extra thermodynamic method. The
`strengths are severalfold: (l) the use of descriptors (1i, u, Es, MR~ and so
`forth) permits data collected from simple organic chemical model systems to
`he utilized for the prediction of biological activity in complex systems; (2) the
`predictions are quantitative witb statistical confidence limits; (3) the method is
`easy to use and is ine~xpensive; and (4) cQnelusions th::u are reached may have
`application beyond the substituents included in the particular analysis,
`The \veaknesses of this method are that (l) there must be parameter values
`availabfo for the substituents in the data set; (2) a large number of compounds
`must be included in the analysis in order to have confidence in the derived
`equations; (3) expertise in statistics and computer use is essential; (4) smaH
`molecule interactions are imperfect models for biological systems; (5) in con~
`trast to chemical reactions i:n which one knows the atoms that interact with
`the reagent, steric effects in biological systems may not be, relevant, sinc.e it is
`often not certain which atoms in the drug interact with the receptor; (6)
`organic reactions used to determine the descriptors usuaJly are studied under
`acidic or basic conditions when all analogs are folly protonated or deproto(cid:173)
`nated, whereas in biological systems tl:.1e drug may be partially protonated; (7)
`since QSAR study is empirical, it is a retrospective technique that depends on
`the pharmacological activity of compounds belonging to the same structural
`type, and, therefore, new types of active compounds are not discovered {i.e.,
`it is a lead optimization technique, not a lead discovery approach); and (8) like
`other empirical relationships~ extrapolations frequently lead to false predic(cid:173)
`tions.
`Despite the weaknesses of this approach it is used widely, and several
`successes in drug design attributable to Hansch an~Jy5;is have been reported. 57
`As pointed out in Section Ul,F,2 of Chapter 3, however, caution should be
`used when applying QSAR methods to racemic mixtures if only one enantio-
`
`

`
`! i
`
`f r
`/. I
`I
`
`iL Drug Development: Lead Modification
`
`37
`
`mer is active. Although Hansch analysis is the foremost method~ there are
`other important statistical approaches that \Vi!I be mentioned briefly.
`
`b, Free and Wilson or de Novo Method. Not long after Hansch proposed
`the extrathennodynamic approach, Free and Wiison58 reporred a general
`mathematical method for assessing the occurrence of additive substituent
`effects, and for quantitatively' estimating their magnitude, It is a method for
`the optimization of substituents within a given molecular framework that is
`based on the assumption that the introduction of a particular substitucnt at
`any one position in a molecule al.ways changes the reiative potency by the
`same amount, regardless of \Vhat other substituents are present in the mole(cid:173)
`cule. A series of linear equations, constructed of the form shown in Eq. (2.18),
`where BA is the magnitude of the biological activity, Xr is the ith substituent
`wlth a value of l if present and 0 if not, ai is the contribution of the ith
`substituent to the BA, and µ. is the overall average activity of the parent
`skeleton, are solved by the method of least squares for the a1 and µ,. All
`activity contributions at each position of substitution must sum to zero. The
`pros and cons of the Free-Wilson method have been discussed by Blankley .59
`Fujita and Ban® suggested two modifications of the Free-Wilson approach on
`the assumption that the effect on the activity of a certain ·substitueut at a
`certain position in a compound is constant and additive. First, they suggested
`that the biological activity should be expressed as log AlA0 • where A and A 0
`represent the magnitude of the activity of the substituted and unsubstituted
`compounds, respectively~ and that ai is the log activity contribution of the ith
`substituent relative to H. This allows the derived substituent constants to be
`compared directly with other parameters related to free energy that are addi~
`tive. Second, they suggested that µ. become analogous to the theoretically
`predicted (calculated) activity of the parent compound of the series. Both of
`these :modifications have been widely accepte-0.
`
`(2.1$)
`
`As an example of the Free-Wilson approach, consider the hypothetical
`compound 2.42.5611 If in one pair of analogs for which R 1, R2, R3
`1 and R4 are
`constant and R 5 is CJ or CH3 , the methyl compound is one-tenth as potent as
`the e:hloro analog, then the Free-Wilson method assumes tbat every Rl
`methyl analog (where R1-R4 are varied) will be' one-tenth as potent as the
`corresponding R5 cbloro analog. A requirement for this approach, then~ is a
`series of compounds that have changes at more than one position. In addition,
`each type of substituent must occur more thao once at each positfon in which
`it is found. The outcome is a table of the contribution to potency of each
`substituent at each position. If the free energy relationships of the ext rather~
`modynamic method are linear or position specific, then Free-Wilson c.akula(cid:173)
`:lions wilt be successful.
`
`

`
`38
`
`2. Drug Discovery, Design, and Development
`
`OH
`I
`CHCH R 1
`2
`
`/
`
`R'-9lt~
`
`R3
`
`R4
`
`2.42
`
`The.interaction model61 is a mathematical modelsimilarto that ofthe.Free(cid:173)
`Wilson additive model with an additional term (exey). to account for possible
`interactions between substituents X and Y.
`
`c. Enhancement Factor. One of the earliest QSAR observations resulted
`froi:n a retrospective analysis of a large number uf synthetic corticosteroids. 62
`Examination of the biological properties. of steroids prepared by the introduc(cid:173)
`tion 9fhalogen, hydroxyl, alkyl, or double bond mPdifications revealed that
`each substituent affects the activity of the molecule in a quantitative sense,
`and almost independently of other groups. The effect (whether positive or
`negative) of each substituent was assigned a numerical value termed the e~
`hancementfactor. Multiplication of the enhancement factor for each substit(cid:173)
`uent by th~ biological activity of Jhe unsubstituted compound gave the po(cid:173)
`tency of the modified steroid.
`
`d. Manual Stepwise Methods: Topliss Operatio11al Schemes and Others.
`Since organic chemists are, by nature, more likely to be intuitive and less so
`rnathema,tical, it was not long before Topliss63 developed a nonmathematical,
`nonstatistical, and noncomputerized (hence, manual) guide to the use of the
`Hansch principles. This method is most useful when the synthesis of large
`numbers of compounds is difficult and when biological testing of compounds
`. is .readily available. It is an approach for the efficient optimization of the
`potency of a lead compound with minimization of the number of compounds
`needed to be synthesized. The only prerequisite for the technique is that the
`lead compound must contain an unfused benzene ring. However, according to
`literature surveys at the time that th1s method was published, 40% of all
`:reported compounds64 cnntain an unfused benzene ring and 50% of drug(cid:173)
`oriented pa.tents65 are concerned with substituted benzenes. This approach
`relies heavily on~ and a values and tD a much lesser degree Es values. The
`methodology is outlined here; a more detailed discussion can be found in the
`Topliss paper.63
`Consider that the lead compound is benzenesulfonamide (2.43, R = H) and
`its potency has been measured in whatever bioassay is being used. Since
`many systems are + 7T dependent, that is, the potency increases with increas(cid:173)
`ing 1T values, then a good choice for the first analog would be one with a
`
`

`
`......
`
`ft Drug Development: Lead Modification
`
`substituent having a +7r value. Since r.4-C! = 0. 71 and o-4-ct = 0.23 (remember,
`1TH "" o-u = 0). the 4-chlorn analog (2.43, R = Cl) should be synthesized and
`tested. There are three possible outcomes of this effort, namely, the 4-chloro
`analog is more potent, equipotent, or less potent than the parent compound. If
`it is more potent. then it can be att.ributed to .a +1T effect, a +u effect, or to
`both. fo this case, the 3 ,4-i:lichloro analog (r.J.4-C!i = l .25, o-3_.i.c11 = 0.52) couid
`be synthesized next and tested. Again, the 3,4-dichloro analog conld be more
`potent, equipotent, or less potent than the 4·ch1oro compound. If it is more
`potent, then determination of whether +1t or +o:: is more important could be
`made by selection next of the 4-SPh analog (1TSPh = 2.32, us1?1> = 0.18) or the
`3~trifluoromethyl-4-nitro analog (1fJ.CF;-4-N02 = 0.60. o-3.cFr4-NO, = l.21).
`
`R-O-S02NHi
`
`2.43
`Al this point a potency tree, termed a 'Topliss decision tree, could be con(cid:173)
`structed (Fig. 2.5), and additional analogs couJd be made. It mmn be stressed
`th~t this analysis was based solely on 1i' and u values, and other factors such
`as steric effects have been neglected.
`If the 3,4-dichloro compound was less potent than the 4~chloro analog, it
`could be that the optimum values of -u and r:r were exceeded or that the 3~
`chloro group has an unfavorable steric effect. The latter hypothesis rouid be
`tested by the synthesis of the 4-trifluoromethyl analog (tr4_cp1 = 0.88, u.i,cp> =
`
`Jt
`
`IM
`4-Cl
`
`IM
`3,4-Cl 1
`
`IM
`4-SPh
`
`I
`IE
`f:lgure 2.5. Topliss decision tree {M, more potent; E, equipotent; l. less potent).
`
`JM
`
`

`
`~f ..... · ..
`
`i
`
`l
`!
`
`4l)
`
`2. Drug Discovery, Design, and Development
`
`0.54), which has 110 3-sub.stitnent but has a high er and intermediate w value.
`If this analog is more potent than the 4-chioro analog, the 4-nitro analog
`(1T4-NO;i = ~0.28, 0'4.No, = 0.78) or the 4-ethyl analog (?T4.Ei = I .02, 0--1-Ei =
`-0. 15) could be synthesized in order to determine the importance of ·rr and fY
`values, respectively.
`What if the 4~chloro analog was equipotent with the parent compound'? This
`could result from a favorable +11 effect counterbalanced by an unfavorable
`(]" effect or vice versa. If the former is the case, then the 4~methyJ analog
`(1T4.Me = 0.56, o-4.Me ""' -0.17) should show enhanced potency. Further en~
`hancement of potency by the 4-methyl analog would suggest that the synthe(cid:173)
`sis of analogs with increasing 1T values and decreasing a values wouJd be
`propitious. lf the 4-methyl analog is worse than the 4-chloro analog, perhaps
`the equipotency of the 44::.hlorn compound was the result of a favorable
`<r effect and an unfavorable rr effect, The 4-nitro analog {·7t4.Nn1 = --0.28.
`U+.No, = 0.78) wouid, then 1 be a wise next choice.
`If the 4-chlorn analog \Vas less potent than the lead, then there may be a
`steric. problem at the 4 position, or increased potency may depend on -'ff and
`-er values. The 3-chlom analog (1r3_c1 = CL71; o-3.a = 0.37) could be synthe(cid:173)
`sized to determine if a steric effect is the problem. Note that the a constant for
`the 3-Cl substituent is different from that for the 4-Cl one because these
`descriptors are constitutive. If there is no steric eJfect, then the 4-methoxy
`compound (1T4.oMe = -0.04, u4-0Me = -0.27) could be prepared to investigate
`the effect of adding a -£.r substituenL An increased potency of the 4-0Me
`substituent would suggest that other substituents with more negative 1T' and/or
`u constants be Hied.
`Topliss63 extended the operational scheme for side-chain problems when
`the group is adjacent to a carbonyl. amino~ or amide functionality. namely,
`-COR, -NHR, -CONHR, ~NJ-ICOR, where R is the variable substit(cid:173)
`uent. This approach is appiicable to a variety of situations other than direct
`substitution on the aromatic nucleus. In this case, the parent molecule is the
`one where R is CH:h and w, u, and E, parameters are used.
`Note that in the Topliss operational scheme, as for other methods in this
`section, the procedure is stepwise. that is, the next compound is determined
`on the basis of the results obtained with the previous one. Three ot.her man(cid:173)
`ual, stepwise methods are mentioned oniy briefly: Craig plots,66 Pibonac.ci
`search rnethod,67 and sequential simplex strategy.68 The Topliss decision tree
`approach evolved from the work of Craig,66 who pointed out the utility of a
`simple graphical plot of 1r versus <T (or any two parameters) to guide the
`choice of a substituent (Fig. 2.6). Once the Hansch equation has been ex(cid:173)
`pressed for an initial set of compounds, the sign and magnitude of the 1r and a
`regression coefficients determine the particular quadrant of the Craig plot that
`is to be used to direct further synthesis. Thus, if both the '1T and u terms have
`
`

`
`41
`
`Pi
`
`lL Drug Development Lead Modification
`
`[+a, ~tr !
`
`0.75
`
`S~NH:i
`•
`
`•CHlSO:i
`
`CCNH;i •
`
`-20
`
`-1.6
`
`-1.2
`
`C~$GONH 1
`~·
`·-0.8
`--OA
`
`11q::
`
`{}
`·~
`
`-0.25
`
`-0CH3
`
`• OH
`
`--0.$0
`
`-!J.1$
`
`~]
`
`Q.4
`
`•. o.s
`SCH3
`•
`
`CH~
`
`12
`
`•
`tt
`
`I -(}, +;r
`
`Flgute 2.s. Craig !)Jot of zy consl<ints versus" values for aromatic sobstltuMts}i& {Reprinted
`by permission of John Wiley & Sons, Inc. from Craig, P. N. {1980}. Jn "Burger's Medidrm.l.
`Chemistry," (M. E. Wolff, ed.J, 4tli od., Part t p. 343. Wiley, New York. Copyright© 1900 John
`Wi!ey & Sons, lnc.J
`
`positive coefficients, then substituents in the upper right-hand quadrant of the
`plot (Fig. 2.6) should be selected fur future analogs.
`The Fibonacci search technique67 is a manual method to discover the opti(cid:173)
`m:um tlf some parabolic function, such as pol.ency versus log P ~ in a minimum
`number of steps. Sequential simplex strategy68 is .another stepwise technique
`suggested when potency depends on two physkochemicatparameters such as
`rr and rr.
`
`e. Bat.cit Selection Methods: Jlau:ilwise Topliss OperationalScheme,, Clus(cid:173)
`ter Analysis, and Others. The inherent problem with the Topliss operational
`scheme described in Section 11,E,J,d is its stepwise nature. Provided that
`pharmacological results can be obtained quickly, this is probably not much of
`a prohJem; however, biological evaluation is often slow. Topliss69 proposed an
`alternative scheme that uses batdnvise analysis of small groups ·Of com-
`
`

`
`2. Drug Discovery, DesJgn, and Development
`
`iilt>le 2.7 Potent:y Order for Various Parameter Dept:ndBnch:s !With pennission from
`Topliss., J. G. {1977), J. Med. Chem 20, 463. Copyright«!> rnn Amar!<:nm Chemical
`Society.]
`
`Substitucnt
`
`rr
`
`211 -
`
`1T1
`
`er
`
`·-~---~,··-----
`
`'lA-Cli
`4-CI
`4-CR,
`,f.!:)CHJ
`H
`
`')
`
`..
`3
`4-5.
`4~5
`
`t-2
`1-2
`3
`4-5
`4-5
`
`2
`4
`5
`3
`
`;,(!'
`
`1i -
`
`17 - ff
`"'
`tr +. l'T
`-er
`2-rr -· er
`··------------------·
`3-4
`5
`I
`f-:2
`J
`3-4
`2-3
`4
`2
`J
`2
`l
`3
`1-3
`l 2
`5
`1
`l
`4
`4
`5
`3
`5
`4
`5
`
`·rr - 3u
`
`5
`3-4
`
`2
`3-A
`
`1-5
`2-5
`2-5
`2-5
`I
`
`~ Unfavorable steric effect from 4-subs!Hutii:m.
`
`pounds, Substituents were grouped by Topliss6fl ace;ording to tr, a\ rtz, and a
`variety of X'lr and ycr weighted combinations_ The approach starts with the
`synthesis of five derivatives, the unsubstituted (4-H). 4-chloro, 3,4-dichloro,
`4~methyl, and 4~methoxy compounds. After these five analogs have been
`tested in the bioassay. they are ranked in order of decreasing potency, The
`potency order determined for ihese analogs is then compared \Vith the rank(cid:173)
`ings in Table 2.7 to determine which parameter or comblnaUon of parameters
`is most dominant. lf, for exa1nple,. the potcucy order is 4,0CHs > 4wCH3 >
`H > 4-Cl > 3,4-Ch, then -rr is t.he dominant parameter. Once the parameter
`dependency is determined, Table 2.8 is consulted in order to discover what
`substituents should be investigated next. 1n the above example, 4-N(Cz}f;)z,
`4-N(CH3)z, 4~NH2. 4--NHC4H!r. 4-0H~ 4-0CH(CH;)z~ 3-CH:,. and 4~0CH1
`W()tdd be suitable choices. The major weakness <~f this approach is that it is
`
`Table 2.8 New Substltuent Selections (With permis