We have created a new Spark database based on PubChem Agrochemical classification, which will be of value to our Agrochemical research and will be a source of different fragments for pharmaceutical research. A paper by Delaney et al.1 compared modern compounds from agrochemical and pharmaceutical research highlighting the similarities between them and opportunities for small molecule research. In addition to this example, we have compared the properties of the fragment databases we created and can see fragments from Agrochemical origins have a significant overlap with fragments from pharmaceutical origins.
A practical example of the benefits of the new agrochemical database for agrochemical research
Topramezone is an herbicide that is sold as a post-emergence treatment initially in maize to control broadleaf and grass weeds. The mode of action is the inhibition of 4-Hydroxyphenylpyruvate dioxygenase (HPPD), and Topramezone is a potent nm inhibitor. Inhibitors of HPPD all contain a substructure that chelates to the active site iron in a bidentate mode (Figure 1). A Spark experiment will be run to identify bioisosteres for this metal chelating warhead. The expectation is that we would find a “triketone-type” fragment as observed in another class of herbicidal HPPD inhibitors such as Mesotrione (Figure 2), as well as in Nitisone which is as an orphan drug treatment for type I tyrosinemia.
Figure 1. Topramezone shown in the bioactive conformation bound to HPPD (PDB: 7CQS). All potent HPPD inhibitors bind to a ferrous iron in a bidentate mode, the phenyl ring stacks between two phenylalanine’s.
Figure 2. Left, Topramezone with the metal warhead circled, which will be the substructure that is being replaced by Spark. Right, Mesotrione, the metal warhead is also circled, this is often referred to as a triketone. Both compounds are weak acids and are shown deprotonated.
The metal warhead of Topramezone was selected as a Spark R-group replacement experiment with the protein as an excluded volume which was searched against ChEMBL, and the Commercial databases, and compared with the results from the new Agrochemical database. Several searches were performed, the first with the default selection (ChEMBL Common, Commercial Very Common and Common databases), secondly with all ChEMBL databases, thirdly searching all Commercial fragment databases. In each case, the top 1000 Spark results were kept for inspection.
Intriguingly, a Mesotrione-like warhead was not identified in the top 1000 Spark results when searching against the ChEMBL or the Commercial databases, it must be noted that the results did contain interesting fragments. When the same search was performed using the Agrochemical fragment database the Mesotrione-like warhead was ranked third (Figure 3).
Encouragingly, this result highlights Spark’s ability to identify bioisosteres and shows the value of searching different databases that contain different fragments. In this case the desired fragment was only found in a domain-specific database, which is not so surprising as it is a specific fragment for a specific protein target. It does however show there is value searching different databases to identify different fragments that can act as bioisosteres.
Figure 3. Spark result with Mesotrione-like warhead, identified by searching against the new agrochemicals fragment database.
Comparing the property overlap between pharmaceutical and agrochemical fragments
In 2006, John Delaney et al. published a review comparing the chemical spaces and measured properties of pharmaceutical compounds and agrochemical compounds. Delaney et al. concluded that generally modern agrochemical compounds are very similar to pharmaceutical compounds in terms of their molecular and measured ADME profiles. Delaney et al. specifically identified herbicide compounds as good starting points for drug design due to their molecular properties.1 With this in mind we compare our Spark databases asking: how similar are the molecular properties and, by virtue, chemical space of the Spark Agrochemicals, CHEMBl_Common, VeryCommon and Common fragment databases (Figure 4).
Figure 4. Mean and median property values calculated across the CHEMBl_Common, VeryCommon and Common and PubChem Agrochemicals fragment databases
In analysing the data reported in Figure 4, a notable overlap in molecular properties can be observed. There is very little difference between the mean and median values for molecular properties that one seeks to typically optimize in a drug design context. This is an interesting finding as it also suggests that the physicochemical characteristics of agrochemical fragments resemble those of pharmaceutical compounds, perhaps more than traditionally assumed. This is pertinent in drug design because it opens the possibility for drug designers to access a broader set of fragment chemistries and apply them to their targets. Such fragments could potentially lead to the discovery of novel therapeutic candidates.
Conclusion
With the introduction of a new agrochemicals Spark database collection, researchers are equipped with the means to another source of chemical fragments with relevant molecular properties to aid in the identification of novel bioisosteric replacements, not only in the context of agrochemical discovery, but also drug discovery.
References
- Delaney, J.; Clarke, E.; Hughes, D.; Rice, M. Modern Agrochemical Research: A Missed Opportunity for Drug Discovery? Drug Discov. Today 2006, 11 (17), 839–845. https://doi.org/10.1016/j.drudis.2006.07.002.
- Dong J., et al. Structural insights of 4-Hydrophenylpyruvate dioxygenase inhibition by structurally diverse small molecules. Advanced Agrochem 2022, 174-181. https://doi.org/10.1016/j.aac.2022.10.002.