Peter J. Taylor

Hydrogen bonding. Part 7. A scale of solute hydrogen-bond acidity based on log K values for complexation in tetrachloromethane

A scale of solute hydrogen-bond acidity has been constructed using equilibrium constants (as log K values) for complexation of series of acids (i) against a given base in dilute solution in tetrachloromethane, equation (A). Forty-five such equations have been solved to yield LB and DB, log Ki=LB log KAHI+DB(A) values characterising the base, and log KAH values that characterise the acid. In this analysis, use has been made of the novel observation that all the lines in equation (A) intersect at a given point where log K= log KAH=–1.1 with K on the molar scale. Some 190 log KAH values that constitute a reasonably general scale of solute hydrogen-bond acidity have been obtained. It is shown that there is no general connection between log KAH; and any proton-transfer quantities, although certain family dependences are obtained. A number of acid-base combinations are excluded from equation (A), and alternative log KAHE values have been determined for such cases. The general log KAH values may be transformed into α2H values suitable for use in multiple linear-regression analysis through the equation α2H=(log KAH+ 1.1)/4.636.

Model solvent systems for QSAR. Part 3. An LSER analysis of the ‘critical quartet.’ New light on hydrogen bond strength and directionality

An LSER analysis of log P for the ‘critical quartet’ of solvent systems has been carried out using, as initial variables, VI for volume, µ2 for dipolarity, and proton donor ∑α and proton acceptor βf scales based on log Kα and log Kβ respectively. A common data matrix and an unprecedented range of functionalities have been employed. By making the analysis stepwise, starting with the simplest solutes and adding more in order of increasing complexity, we have been able to identify hitherto unrecognised variables and ‘fine-tune’ established ones in such a way as to derive self-consistent proton donor and acceptor values applicable to the whole range of solvent systems. By this ‘LSER in reverse’ we have established, inter alia, the following new facts: (a)βf possesses a constant effective zero whereas that for ∑α is solvent-sensitive; (b) a new term nβf is required for acceptor solutes with two or more available lone pairs; (c) when neither lone pair is available, the acceptor strength of carbonyl is sharply reduced; (d) a second term specific to NH2 is required for ∑α in alkane and chloroform; (e) there is mutual shielding of XH and one lone pair in structures such as CO2H and CONH2; (f) ureas and other structures with parallel NH functions are proton donors of exceptional strength; (g) the acceptor ability of dipolar bases (PO and SO) varies with the solvent system.Cooperativity in solute–solvent bonding exists but takes complex forms, and does not appear strong enough to account e.g. for the hydrogen bonding properties of bulk water and the alcohols, for which mass action appears a likelier explanation. We present evidence (see Appendix) that mass action will most probably explain certain well known anomalies in the apparent proton acceptor ability of water as revealed by partitioning studies.The present results throw new light on previously anomalous octanol–water log P values and can predict f-values for other solvent systems. Most importantly, they provide new information not only on the strength of hydrogen bonding for more than 60 functional groups, but also on its directionality: we are able to predict, with reasonable certainty, which XH groups and lone pairs are actually available for bonding. This information is applicable to water, other solvents, and by implication to the biophase, so should find direct and immediate use in rationalising and quantifying drug–receptor interactions.

Hydrogen bonding. Part 9. Solute proton donor and proton acceptor scales for use in drug design

Hydrogen bonding equilibrium constants have been measured for a large and varied selection of proton donors against a common acceptor (N-methylpyrrolidinone) and of proton acceptors against a common donor (4-nitrophenol). Together these have been used to create the log Kα and log Kβ scales of proton donor and acceptor ability which are explicitly targeted to the needs of the medicinal chemist in the context of potential drug–receptor interactions. To this end they have been measured in 1,1,1-trichloroethane, a solvent never before used for hydrogen bonding studies but whose high dipolarity is considered a much better model for real biological membranes than the very non-polar solvents that have previously been employed. It is shown that this solvent imposes significant ranking changes on the solutes, since the charge transfer element in hydrogen bonding is reinforced at the expense of the purely electrostatic component. Nevertheless it is possible to scale previous data in such a way that over 80 functional group log Kα and log Kβ values become available to the medicinal chemist (Table 4). In addition, data are given for a large number of parent heterocycles, most of which have never before been studied. We note that heterocycles are uniquely able to ‘fine-tune’ these scales, so providing at least one justification for their special interest to the medicinal chemist.In addition to equilibrium constants we have measured the spectroscopic quantities ΔνCO(for donors) and βsm(for acceptors). On various lines of evidence we suggest that these are enthalpy-related quantities and, following previous arguments, may function as alternative parameters suitable for use by the medicinal chemist under conditions of severe steric constraint.Cross-comparisons of these data allow conclusions to be drawn which considerably illuminate the factors that influence hydrogen bond strength, and some of which have no precedent. A selection follows. Where a level comparison can be made, the donor order is OH > NH > CH and the acceptor order is N > O > S. However, within each category there are various sorts of family relationship. For example, phenols and alkanols lie on separate lines of log Kαvs. pKa, and a similar separation for log Kβ is shown by 5- and 6-membered ring heterocycles. By contrast, OH and NH donors show a single relation between log Kα and ΔνCO, negative deviations from which are satisfactorily accounted for in terms of steric and stereoelectronic factors. The most important of the latter is lone-pair repulsion: ‘α-effect’ heterocycles are anomalously strong acceptors, whereas certain classes of donor, notably sulphonamides and carboxylic acids, are much weaker than would be expected from their pKa values. More subtle anomalies attach, inter alia, to heterocycles as donors, CH donors generally, and amines and sulphonamides as acceptors; all however can be rationalised.The extremes of both scales are charted. Alkyl thiols and amines are negligible as proton donors; correspondingly, π-donor hetero-atoms as e.g. in esters and amides are negligible acceptors. At the opposite extreme, heterocycles such as tetrazole and 4-quinolone figure prominently. Based on these results, some structural criteria are suggested that might lead to the synthesis of stronger proton acceptors than any so far known.

Model solvent systems for QSAR. Part 2. Fragment values (‘f-values’) for the ‘critical quartet’

A data matrix has been prepared of log P values for 103 compounds distributed across four highly contrasted solvent–water partitioning systems: the ‘critical quartet’ of octanol (amphiprotic), alkane (inert), chloroform (proton donor) and propylene glycol dipelargonate (PGDP; proton acceptor). Here ‘alkane’ is defined as the straight-chain sequence from hexane to octane and (possibly) higher; it is shown that cyclohexane is out of line. In principle, these log P values can now be used to construct a comparative table of fragment values (f-values) for all four systems. In practice, those for non-polar substituents must first be established. Here the key quantity is f(CH2). This has been re-determined, and in the process its variability rationalised, for 24 water-saturated solvent systems; here the key factors (dry solvents are different) turn out to be the molarity, in the organic phase, of water and the solvents own functional group. There results an almost complete data matrix of 82 f-values for all four solvents, about 25% of which are derived from the linear solvation energy relationship (LSER) equations of Part 3.7 It is shown that these four sets are very distinct, a fact that misleading statistical treatments can easily disguise. How the medicinal chemist might use these contrasting data sets is critically discussed, with particular reference to the rationalisation of biological selectivity.

Theoretical studies of vibrational frequency shifts upon hydrogen bonding. The carbonyl stretching mode in complexes of formaldehyde

The shifts in the vibrational frequencies of formaldehyde and water upon formation of the hydrogen-bonded complex have been calculated using ab initio molecular orbital wavefunctions and STO-3G, 3-21G, 6-31G and 6-31G** basis sets, and are compared with experimental values. The shifts in the formaldehyde frequencies are given best in the STO-3G basis, whilst the 6-31G** basis is needed to predict the changes in the water frequencies accurately. For a series of complexes involving formaldehyde studied at the STO-3G level, a linear relationship is found between the calculated hydrogen bond strength and the shift in the carbonyl stretching frequency.

Hydrogen bonding. Part 1.—Equilibrium constants and enthalpies of complexation for monomeric carboxylic acids with N-methylpyrrolidinone in 1,1,1-trichloroethane

A novel calorimetric method has been derived for the simultaneous determination of equilibrium constants and enthalpies of complexation of monomeric carboxylic acids with bases in an inert solvent. The method requires a knowledge of the corresponding equilibrium constants and enthalpies for the monomer/dimer equilibrium in the same solvent. Both sets of K° and ΔH° values have been obtained for a number of carboxylic acids (and some other hydrogen-bond donors) in 1,1,1-trichloroethane at 298 K, using N-methylpyrrolidinone as a standard base. For the first time, it is possible to evaluate the relative hydrogen-bonding strength of monomeric carboxylic acids and other hydrogen-bonding species in an inert solvent. It is shown that unactivated carboxylic acids are no stronger than simple phenols: equilibrium constants for hydrogen bonding towards N-methylpyrrolidinone are acetic acid (109), benzoic acid (118) and phenol (137). It is further shown that the monomeric carboxylic acids are ca. 20 times as strong as the dimeric acids towards N-methylpyrrolidinone (NMP) in 1,1,1-trichloroethane, with respect to the formation of the species RCO2H·NMP in each case.

The tautomerism of 1,2,3-triazole in aqueous solution

Two circumstantial but independent arguments both confirm that the 2H-tautomer of 1,2,3-triazole is favoured in aqueous solution by a factor of ca. two. Lone-pair repulsion is the probable cause; evidence from the solvent dependence of the tautomeric ratio is discussed.

Analysis of hydrogen-bond complexation constants in 1,1,1-trichloroethane: the α2Hβ2H relationship

Hydrogen-bond complexation constants determined by Taylor and co-workers using 1,1,1-trichloroethane (TCE) solvent have been analysed through the α2Hβ2H relationship; α2H and β2H are the solute hydrogen-bond acidity and basicity parameters obtained from complexation constants in tetrachloromethane. Constants for three alcohol/N-methylpyrrolidinone complexations have been determined in TCE, and if these are used instead of the original alcohol/N-methylpyrrolidinone complexation constants, a good relationship is obtained, eqn. (i). The slope in eqn. (i) is smaller than that for the α2Hβ2H relationship in tetrachloromethane, but the intercept is the same.Eqn. (i) has been used to obtain 25 new α2H values for acids; these include acetanilides, sulfonamides, triazoles and tetrazoles. The latter two types of compound have very large α2H values; that for 5-phenyl-1,2,3,4-tetrazole (0.88) being near the value for dichloroacetic acid (0.90). Values of β2H for 31 hydrogen-bond bases have also been calculated using eqn. (i). These include bases with heterocyclic moieties to which β2H values had not previously been assigned, e.g. oxazole, isoxazole, triazoles and a tetrazole.

On the calculation of tetrahedral intermediate pKa values

The procedure of Fox and Jencks for calculating tetrahedral intermediate pKa values, as ΔpKa=ρI∑σI; relative to that of some defined amine or alcohol, is re-examined in the light of more recent estimates for σI. Using those of Charton, which are explicitly tuned to aqueous or near-aqueous conditions, we derive a value of ρI=–9.1 ± 0.4 for the effect of substituent X on probe Y for a one-carbon separation (X–C–Y). Additionally we derive, for X–C–C–Y, a value of ρI=–4.4 ± 0.4.We also examine the possibility of assigning σI values to charged substituents. It is shown that this approach can be made to work under strictly defined conditions, and results in a self-consistent set of σI± values that may be used in the present context.

Hydrogen bonding. Part 3.—Enthalpies of transfer from 1,1,1-trichloroethane to tetrachloromethane of phenols, N-methylpyrrolidinone (NMP) and phenol–NMP complexes

Enthalpies of solution of seven phenols and of NMP have been determined in 1,1,1-trichloroethane and in tetrachloromethane at 298 K. Combination with our previously determined enthalpies of complexation, ΔH°, of the phenols with NMP leads to enthalpies of transfer of the hydrogen-bond complexes ArOH ‥ NMP from 1,1,1-trichloroethane to tetrachloromethane. The substituent-dependent behaviour of values of ΔH° in the two solvents arises exclusively from the effect of the solvents on the reactant phenols and not on the complexes themselves. This finding confirms the suggestion that the ΔH° values in 1,1,1-trichloroethane contain a contribution from association of the phenols with the solvent itself, probably of the dipole–dipole type.