“Druglike” space – an artefact of the reactions we (medicinal chemists) use?

Scientists from AstraZeneca have recently carried out an analysis of the impact of organic synthesis on medicinal chemistry programs and revealed that, with a few exceptions, the most common reactions used in 1984 are still used in 2014, and that the chemical space generated with those few more frequently used reactions is composed of structurally similar compounds and therefore maybe biasing the drugs that are emerging.

In the article, Brown and Boström compare the most frequent reactions in Medicinal Chemistry in 2014 versus 1984 (see Figure 1) and found out that only a few newer reactions have taken a space in the top 20 list, such as the Suzuki and Buchwald cross-coupling reactions. While the latter was first published in 1994, and therefore could not have been used a decade before, the Suzuki-Miyaura reaction was first published in 1981 but the impact of this reaction was not seen in 1984. Similarly, newly developed synthetic reactions have yet to show an impact in recent drug discovery programmes.


Figure 1. Occurrence of a particular reaction, plotted as percentage of which it shows up in at least one manuscript (Figure taken from DOI: 10.1021/acs.jmedchem.5b01409)

Instead, comparing the reactions used by medicinal chemists to those used by chemists working towards the synthesis of natural products shows a completely different picture (see Figure 2). While the former tend to make amide bonds, aryl-aryl bonds, and amino-aryl bonds, the latter concentrate on functionalising oxygen atoms, set stereocentres and make carbon-carbon bonds.

Although carbon-carbon bond formation is a common practice, the type of reactions used by medicinal chemists and natural product chemists differs significantly. Medicinal chemists tend to used Suzuki coupling reactions while in natural product production it varies with aldol, Wittig and Grignard reactions.


Figure 2. Occurrence of a particular reaction type plotted as percentage of which it shows up in at least one medicinal chemistry manuscript versus natural product papers (Figure taken from DOI: 10.1021/acs.jmedchem.5b01409)

The authors question whether the type of reactions selected by medicinal chemists is done out of convinence (e.g because of efficiency and chemoselectivity of the reaction), therefore leading to libraries of compounds  with similar shape, or whether it is indeed there where the drug space is more abundant. To address this question they examine the frequency of biphenyl fragments (normally achieved through Suzuki reaction) in an AstraZeneca collection and found that over time there has been a 6-fold increase on the appearance of this fragment.  Using an in-house database (IBEX) they further examined the possible substitution patterns of mono– and disubstituted biphenyl structures.   For monosubstituted biphenyl fragments it was found that the para substituion was preferred while for the bisubstituted compounds it was the paraortho arrangement (Fig. 3). These preferences result in a high density of linear and disk shape molecules (see Figure 4 green and blue dots), whether the less frequent substitution patterns lead to more diverse molecular shapes (see Figure 4 red dots).


Figure 3. Frequency population of various biphenyl regioisomers in the IBEX (Figure taken from DOI: 10.1021/acs.jmedchem.5b01409)


Figure 4.Population analysis of representative biphenyl compounds illustrating the geometrical diversity of para-para (green), meta-para(blue), and ortho-ortho (pink) compounds (Figure taken from DOI: 10.1021/acs.jmedchem.5b01409)

Brown and Boström point out that, despite the reactions being used by medicinal chemists today still rely on chemistry discovered decades ago, scientific and technological innovations are still of great influence in medicinal chemistry laboratories. Automation, microwave and supercritical fluid chromatography are just some examples of technological advances  and recent reaction improvements have provided advantages to the classical reactions conditions discovered decades ago. However, the impact of recently developed methodologies, such as ring-closing metathesis, C-H bond activation, selective fluorination, biocatalysis etc… is still to be seen in medicinal chemistry laboratories.


Blog written by Carol Villalonga-Barber

Facile strategy for the creation of complex and diverse compounds

High-throughput screening (HTS) of synthetic chemical libraries, containing mainly small molecules, is widely used in drug discovery programmes, both in industry and academia.

HTS has provided many drug leads, but mainly for biological targets that can be modulated by low molecular weigh and planar compounds. For more complex biological targets, HTS will fail due to the nature of the composition of the screening library. A recent study (J. Med. Chem. 2011, 6405) has shown that medicinal chemists have been synthesizing, over the last 50 years, compounds with lower than ideal Fsp3 (fraction of sp3-hybridised carbons) values and higher than ideal ClogP values, the former attributed to the increasing ease of sp2-sp2 coupling reactions. Therefore there is an interest in creating new libraries of complex compounds with better “druglike” features.

Hergenrother and co-workers (Nature Chemistry, 2013, 195) describe a ring-distortion strategy to rapidly (≤5 synthetic steps) generate collections of complex and diverse small molecules from readily available polycyclic natural products. An important consequence of starting with natural products is that all the intermediates generated are complex structurally and worth of inclusion in the final library.

They demonstrate this strategy for three complex natural products, gibberellic acid, adrenosterone and quinine using combinations of ring-cleavage, ring expansions, ring-fusions and ring rearrangements reactions (Fig 1-3).

cv4Figure 1. Ring-distortion approach on gibberellic acid.

cv5Figure 2. Ring-distortion approach on adrenosterone.

cv6Figure 3. Ring-distortion approach on quinine.

The average Fsp3 values for Hergenrother compounds was found to be 0.59, considerably higher than the 0.23 average found for a ChemBridge commercial collection of 150,000 compounds, while the ClogP was 2.90, 1.1 log units lower that that in the commercial collection, corresponding to a 12-fold reduction in hydrophobicity. Moreover, a chemoinformatic analysis (Tanimoto coefficients) revealed very low similarity between all of the compounds synthesized in this way which is a much superior derivatisation strategy than the conventional modification of peripheral functionalities.

Screening this library should definitely be of great interest to medicinal chemists.