Why is diazepam a weak base

Figure 4 shows a set of representative bases containing various heterocycles and amines. In all, compounds contained a single base making up just under half of the total set analyzed. Figure 5 shows the distribution of base p K a values ranging in value from 0.

Chart showing nine bases with a range of p K a values. In each case the basic group has been highlighted with an arrow. Diagram showing the p K a distribution of compounds containing a single basic group. Each group of columns contains a comparison of the entire set of single bases and those from the CNS and non-CNS subsets.

Compounds were binned into 1 log unit ranges as per Figure 3. Indeed, there were no bases with a value above Once again the majority of compounds had a p K a above 7 and mostly consisted of amines.

The distribution for the non-CNS class closely matched the overall pattern found for the entire dataset with a peak in p K a values at around 9. In order to analyze the distribution of simple ampholytes i. Figure 6 illustrates the range of isoelectric points for both the ordinary and zwitterionic ampholytes. While no clear pattern emerges this may be a reflection of the limited number of compounds 65 available for this analysis.

The larger number of ordinary ampholytes at the high end of the scale represent simple phenols with alkylamine side chains e. If these compounds are left aside, those that remain tend to have isoelectric points between 3. Histogram comparing the isoelectric points of both ordinary and zwitterionic ampholytes. In this case the frequencies of the distributions were shown to reflect the differing number of ordinary ampholytes 44 compounds and zwitterionic ampholytes 21 compounds.

For the CNS class there were 13 simple ampholytes which made up only 7. Of these 13 compounds there were six opioids and six benzodiazepines all of which were ordinary ampholytes. In contrast, the non-CNS subset contained 52 ampholytes comprising 20 zwitterions and 32 ordinary ampholytes.

No doubt the predominance of ordinary ampholytes in the CNS class reflects the neutral character of these compounds at their isoelectric point where neutrality would favour CNS penetration. One concern over the analyses conducted in this study may be the choice of datasets used. This is a problem that plagues any analysis of drug sets that aim to tease out trends in physicochemical characteristics.

The set employed should of course be representative of drugs as a whole to enable reasonable conclusions to be drawn. To look at the proportion of ionizables the WHO essential medicines list 48 was used which represents a small pharmacopoeia for priority health care needs.

It is overrepresented in certain drug classes e. Nevertheless it is a well thought-out list covering the majority of therapeutic classes. In contrast, the WDI dataset used by Comer and Tam 29 , 30 consisted of 51, compounds and could be viewed perhaps as a master list of drugs. The WDI, however, includes pesticides, herbicides and compounds that did not reach the market place. Given our desire to be representative of drugs it is not an ideal set and may be considered too encompassing.

Our analysis therefore of the proportion of compounds that are ionizable is very dependent on the dataset used and provides results specific to that set. Another option is to examine all the drugs used commercially around the world such as those listed in Martindale This contains over drug monographs and an analysis based on this set would be an onerous task.

The obvious alternative is to choose a smaller set that has undergone an evolutionary process to select useful therapeutic substances e. Until such time that an agreed set of compounds can be selected to determine how many are ionizable the numbers generated here using the WHO list A more interesting analysis might be where strict criteria are used for compounds to be included in a survey. For small organic substances this would give a better indication of the proportion of compounds possessing an ionizable group.

The Williams list of compounds 33 could also be scrutinized in the same manner as the WHO essential medicines list. It is however, an extensive set of substances and represents a wide range of therapeutic classes. Once again better and more recent sets could be devised for this study and the Williams set was selected as a useful representative set and for the large number of compounds it contained.

As mentioned above this aspect of the study is being addressed in future work in these laboratories using the compounds listed in the AHFS Drug Handbook Until such time that these larger and more recent data sets are analyzed this present study provides an interesting insight into both the proportion of ionizable substances and the distribution of p K a values.

The power of the present analysis is to flesh out the bones to this simplistic description and provides a starting point for discussing p K a distributions. In particular, the apparent biphasic distribution of acid p K a values needs to be investigated further. Another important aspect to this research has been the scrutiny applied to CNS compounds.

While, there is a general understanding concerning the principles behind the distribution of acid and base p K a values for CNS drugs, this has not been well documented or presented in the literature. For example, it is known about the paucity of CNS compounds with acid p K a values below 4. Also recognized is the sensibility of these values as charged substances do not easily cross the BBB. The cutoff values described by Fischer and coworkers 32 concur with the observations presented here, although only one compound had an acid p K a below 6.

The important aspect of this present study was to outline the distributions themselves to demonstrate the spectrum of p K a values. Indeed, the overall implication is that this is valuable information when contemplating the properties needed for a drug or sets of screening compounds.

The utility of the distributions described here may be applied to third party supplier databases for purchasing decisions regarding screening compounds. Either the ratio of ionizable to neutral compounds could be applied or the p K a distributions could be used in the selection process.

One thing that needs to be borne in mind is that the work described in this study has emerged from an analysis of drugs. Given that current screening efforts are oriented to lead-like molecules 15 then the distributions need to be considered in this light.

Certainly an analysis of an ideal screening set of lead-like compounds would yield the appropriate data. In the absence of this we need to look at the guidelines suggested for lead-like character. In other words there is scope for chemists to take a small molecule with reasonable activity and enter this into rounds of optimization for activity, selectivity and biopharmaceutical properties.

The physicochemical criteria listed above are very simple, however p K a and logD are not considered. Perhaps a simple ratio of ionizable to non-ionizable compounds needs to be suggested e. Furthermore the makeup of the ionizables also needs to be considered by selecting compounds with single acids, single bases and ampholytes, in approximately the ratios outlined in Table 2.

More complicated combinations of acids and bases or those with 2 or more acids and bases should be kept to a minimum. These suggestions are purely speculative and are open to debate; suffice to say that the compounds should contain a mix of neutral and ionizables in roughly the ratios seen for drugs as well as allowing chemists the possibility of adding further ionizable groups to enhance activity and biopharmaceutical characteristics as part of the optimization process.

Ionizable groups on drug molecules have two principal functions. The first is to modify overall polarity, which in turn controls other physicochemical properties, such as aqueous solubility or hydrophilicity.

The second is to provide functional groups that can interact with target macromolecules in specific ways. Organic chemists, on the other hand, do not necessarily consider ionizable groups as first priority groups to include on a novel compound. A chemist, for ease of synthesis may prefer to work with non-polar compounds that are soluble in organic solvents. Another human consideration is the simplicity of the chemistry.

Straightforward synthetic schemes will no doubt predominate to reduce the number of steps required. Given that ionizable groups often require protection means that additional synthetic steps are needed and introduces a further level of difficulty. Taking all this together suggests that organic compounds made to date will largely be lacking in ionizable groups.

Furthermore, many of the third party suppliers need a large number of new substances for their catalogues which means that a high throughput is required from their chemists. High throughput will be a driver for simpler chemistry and, using the argument above, will result in compounds lacking ionizable groups. Of course, this trend has been identified and is being specifically addressed for compounds with utility in medicinal chemistry.

Medicinal chemists also follow the principles of organic chemistry and prefer to introduce polar ionizable groups in the latter stages of a synthesis e.

The last step of a synthesis can also be engineered to be one that can introduce diversity to generate a set of analogues. Third party screening compound suppliers, however, obtain a proportion of their catalogue from organic chemists rather than medicinal chemists. An overriding question fundamental to this study concerns the p K a distributions themselves. Two separate influences will ultimately shape these findings. The first is chemical in nature concerning the functional groups that comprise the acid and base moieties.

If we took the universe of organic compounds a good representative subset might be the organic compounds contained in the CAS collection and produced p K a distribution plots then it would be possible to see how drugs compare.

Similar arguments could be directed at basic compounds and that the distributions we observe for drugs are a function of the regularly seen groups used in these compounds.

Certainly, toxic functional groups will be very limited in the Williams set 33 and this may also affect the p K a distribution. The second driver for the p K a distributions is biological in nature and is affected by membrane properties and the drug targets themselves.

Our knowledge of p K a distributions for a number of functional groups is quite reasonable but not when these are considered collectively. Presumably the pK a value is a quantity which does not have a smoothly distributed continuum of values, but is necessarily multimodal because of the types of functional groups that exist in organic chemistry. In that sense, it is unlike logP, which has a much more broadly distributed set of values.

This is a research area that will no doubt develop as larger populations of compounds are studied. The task of identifying acids and bases in a database is a readily achievable task.

A more difficult procedure is to estimate the p K a values for these compounds. With regard to accuracy we preferably seek to predict within one log unit of the measured value. A variety of computational approaches are available and this topic was reviewed recently by Wan and Ulander 7. A number of methods are used within the commercial packages e. Typically, a molecule is fragmented and the p K a of the functional group is estimated by referring to a database of values with associated QSAR equations.

Artificial neural network methods have also been used to estimate p K a and the software available from Simulations Plus is one such example The other primary method of estimating p K a values is through quantum mechanical techniques. The advantage here is that they can adapt to new chemical classes and do not necessarily need prior examples within the algorithm. In each case, and to differing degrees, estimates can be complicated by conformational flexibility, solvent handling, conjugated systems and a lack of relevant examples.

The needs of the pharmaceutical industry are challenging as they regularly explore novel structural scaffolds to enter new patent territory. If the software requires prior examples of a functional group or scaffold then accuracy may be compromised. For the purposes of characterizing a database, speed of calculation is a priority and may take precedence over accuracy. Among the considerations are problems such as conformational flexibility, internal hydrogen bonding, solvent effects and multiprotic influences 7.

Fortunately, several groups are working on better prediction methods and this will ultimately influence how we undertake research for new medicines. This study has begun to explore the overall composition of drugs with regard to the proportion of those compounds containing an ionizable group. Within the WHO essential medicines list Other estimates give this number as anywhere between Analysis of Williams collection of drugs 33 has led to a description of the relative proportions of compounds containing acidic and basic functionality.

More importantly, the distribution of p K a values has been outlined in detail for the first time. Two clear findings emerged upon examination of the distributions particularly when a distinction was made between CNS and non-CNS drugs. Firstly, acid p K a values for CNS drugs rarely fell below 6. From an ionization viewpoint these observations are entirely reasonable when considering the nature of the BBB and the passage of charged substances across membranes.

As such, these observations consolidate current wisdom in the area and open the way for larger collections to be compared to these distributions. Without doubt p K a is of paramount importance to the overall characteristics of a drug and has considerable influence on biopharmaceutical properties. Current trends indicate that future research is placing an increased focus on p K a with the advent of high throughput measurement techniques and improvements to computational prediction software 7.

By taking p K a into account allows the researcher to begin ADME profiling early in the discovery process. Moreover, with large collections of compounds such as corporate databases, third party supplier offerings and virtual sets of compounds e.

If these differ largely from the observations outlined in the current study then it allows the opportunity to amend synthetic directions or screening compound selections. The drive to consider the physicochemical properties of drugs to understand biopharmaceutical characteristics began many years ago e.

This has fundamentally changed how discovery work is undertaken and was oriented to improving the efficiency and productivity of pharmaceutical companies. Likewise, the need to explore p K a will begin to influence how we work. The findings presented here go some way to understanding the distribution of p K a values and further guidelines will evolve as larger datasets are analyzed.

The Williams 33 dataset has been provided as supplementary material. The author thanks Drs Richard Prankerd and David Chalmers for their insightful discussions and valuable suggestions.

Dedication: Dedicated to the memory of Professor Martyn Ford. Since this article was written Lee et al. Lee P. Comparisons of predicted against measured p K a values for a set of drugs gave a root mean square error of 0. The program is also capable of running in batch mode and may be extremely useful for characterizing large data-sets of compounds.

National Center for Biotechnology Information , U. Journal List Perspect Medicin Chem v. Perspect Medicin Chem. David T. Author information Copyright and License information Disclaimer. Correspondence: David T. Manallack, Email: ua. This article has been cited by other articles in PMC. Associated Data Supplementary Materials The Williams 33 dataset has been provided as supplementary material.

Abstract The acid-base dissociation constant p K a of a drug is a key physicochemical parameter influencing many biopharmaceutical characteristics. Keywords: pK a , dissociation constant, distribution, drugs, absorption, ADME, bioavailability, drug discovery, pharmacokinetics, acids, bases, ampholytes.

Introduction An awareness of the influence of the acid-base dissociation constant, p K a , on the biopharmaceutical properties of drugs and chemicals has long been established within the pharmaceutical and chemical industry.

Drug-likeness In recent years there have been numerous studies exploring methods to improve the efficiency of the early stages of new medicines research. Open in a separate window. Figure 1. Results a Acid and base proportions The proportion of acids and bases in the Williams 33 dataset of compounds was determined by reviewing the p K a data and summing the number of compounds containing a single base, single acid, and so forth. Table 1. Table 2. Figure 2. Figure 3. Figure 4. Figure 5. Figure 6.

Discussion Overview of findings One concern over the analyses conducted in this study may be the choice of datasets used. Application of findings The utility of the distributions described here may be applied to third party supplier databases for purchasing decisions regarding screening compounds.

Perspectives and future directions Ionizable groups on drug molecules have two principal functions. Conclusion This study has begun to explore the overall composition of drugs with regard to the proportion of those compounds containing an ionizable group.

Supplementary Material The Williams 33 dataset has been provided as supplementary material. Click here to view. Acknowledgments The author thanks Drs Richard Prankerd and David Chalmers for their insightful discussions and valuable suggestions.

Physicochemical profiling: overview of the screens. Drug Discov Today: Technologies. Physicochemical profiling solubility, permeability and charge state Curr. Kinetics of distribution and adipose tissue storage as a function of lipophilicity and chemical structure.

Drug Metab. Quantitative structure-pharmacokinetic relationships amongst phenothiazine drugs. The pharmacokinetics of antiarrhythmic agents in pregnancy and lactation. Expert Opin. Drug Metab Toxicol. J Pharm. Recent advances in computational prediction of drug absorption and permeability in drug discovery. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Del Rev. The Design of Leadlike Combinatorial Libraries.

Angew Chem. Molecular properties that influence oral drug-like behavior. Curr Opin. Drug Discov. Drugs, leads, and drug-likeness: an analysis of some recently launched drugs. Is there a difference between leads and drugs? A historical perspective.

J Chem. For an excuse to play with Illustrator, this relationship can be presented as a series of translucent fluid-filled tubes. This diagram depicts the effect of a change in pH on the lipid solubility of a weak acid. Seeing as many drugs are either weak acids or weak bases, they will either be charged or uncharged in solutions with different pH.

Generally speaking the pH of extracellular fluid is always going to be within some decimal fractions of 7. Weakly basic drugs with a pKa closer to 8 will usually be lipid-soluble and will therefore find it easier to negotiate the barrier membranes on their way to their target.

However, it does not describe all possible cases. For instance, zwitterions hermaphroditic neutral molecules with both positive and negative polar groups penetrate lipid bilayers by presenting themselves "side-on" to the hydrophobic membrane, thus appearing as neutral non-polar molecules while they pass.

It is thought that fluoroquinolones gain intracellular access in this manner Cramariuc et al, Moreover, some ionised substances are present in such high concentrations that they are able to cross the lipid bilayer purely by the brute force of their concentration gradient the classic example of this is water: the concentration of water in pure water is Charifson and Walters present an excellent graph reproduced below with no permission whatsoever to demonstrate the distribution of pKa values across the commonly used substances.

They selected all available drugs in ChEMBL and DrugBank, provided they were made up of at least 10 "heavy atoms", had a molecular weight under and contained a reasonably conventional bunch of elements no lanthanides or anything. The final data set ended up being a collection of drugs. The authors went further yet by analysing the pKa distrubition according to drug class, route of administration, clearance mechanisms, and so forth.

Beautifully colourful graphs were produced. The curious exam candidate with infinite time resources is directed to the original paper for more details, but the basic findings consisted of several broad trends:.

Generally, it was found that there are more basic drugs among those agents which target membrane receptors and transporters, whereas those which target enzymes and ion channel tend to be more neutral. Trapping effects take place when drugs cross a lipid membrane and enter an area with a significantly different pH to the one they previously occupied.

The change in pH may suddenly render the drug more ionised and therefore less lipophilic. Unable to cross the membrane in the opposite direction, ionised drug molecules will become concentrated in this ionising solution, a phenomenon known as "ion trapping". The use of this in toxicology is probably the most interesting clinical application of the concept.

It is a method of increasing drug clearance which depends on the premise that alkaline urine favors excretion of weak acids and acid urine favors excretion of weak bases. In this fashion, we are instructed to alkalinise the urine to promote the excretion of weak acids such as salicylate and urate. Its not just urine. Also, acidic environments of abscesses can interfere with polarity of local anaesthetics, making them less lipid soluble and thus less effective.

Again for no reason other than amusement, the author will conclude with a list of body fluids and their respective pH values so that inquisitive minds can create thought experiments exploring the ion trapping effects which might take place at the interface of blood, saliva, gastric acid, semen and vitreous humour.

Depending on who is sampled and which textbook you read, these values may be slightly different. Charifson, Paul S. Patrick Walters.