Introduction

Insulin-regulated aminopeptidase (IRAP) is a transmembrane protein critical for a myriad of unique and independent biological processes, including peptide regulation, insulin-responsive translocation, cardiovascular regulation and cognitive function. In fact, the latter of these actions place IRAP at the epicentre of ongoing Alzheimer’s research, as small molecule inhibitors of IRAP have the potential to enhance cognitive function in Alzheimer’s patients by stabilizing neuropeptides and improving synaptic plasticity. Recently, Mpakali et al.,¹ published a study targeting the enzymatically active open conformation of IRAP and focused on the discovery of small molecule IRAP inhibitors using Cresset’s 3D virtual screening solution, Blaze™ .

In this article, we highlight the successful use of Cresset’s virtual screening technology to the real-world application of small molecule IRAP inhibitor identification. We discuss how Cresset’s ligand-based 3D virtual screening technology was used to screen millions of compounds in silico, retrieving good hit rates.

Virtual screening in 3D field descriptor space with Blaze

During screening campaigns, in silico screening against pre-enumerated chemical collections containing millions of compounds is routinely performed. Typically, it is common practice to rely on the computationally less expensive 1D or 2D fingerprint descriptors to evaluate the compounds’ similarity. However, such approaches are limited in offering structurally novel results due to the simple neglect of any 3D electrostatic and shape field information known about the query. In specific, describing molecules on the basis of connectivity and substructure alone is an oversimplification that often fails to circumvent chemotype bias.

Blaze overcomes this issue by screening in 3D electrostatic and shape field space. By utilizing the XED-computed field point pattern,³ which by virtue encompasses all of the relevant information for the key interactions a ligand can make with a protein. Such descriptors are ‘chemistry agnostic’, assessing candidate compounds on the basis of how the protein ‘sees them’.

Compounds with high similarity, in terms of the electrostatic and shape field point patterns, are expected to map closely to the bioactivity of the query molecule. To that end, the Blaze workflow is as follows:

1. The 3D bioactive conformation of the query molecule is supplied, and its field point pattern computed.
2. From this, Blaze executes a multi-tiered search (Figure 1) over conformationally enriched databases of commercial (“off-the-shelf”) compounds (> 30M compounds).
3. The field point pattern similarity between the query molecule and the compounds within the database is calculated on the basis of a 50% electrostatic potential 3 and 50% shape 4 contribution to compute an overall similarity score. Although, in practice, these parameters can be adjusted for further diversity.
4. Hits are then identified as compounds with high electrostatic and shape field point pattern similarity to the query.
5. Blaze returns a ‘hit list’ ranked by 3D electrostatic and shape similarity to the query, differentiating this method from the binary “pass/don’t pass” output typically observed when performing pharmacophore-based virtual screening.

Figure 1. The Blaze 3D virtual screening workflow

Blaze retrieves hits against IRAP and instigates the start of a medicinal chemistry optimization campaign

In the work of Mpakali et al., the query molecule for the Blaze search was the DG025 transition state inhibitor (sequence 2X0-4L8-KHHAFSFK) in complex with IRAP (PDB: 4Z7I, Figure 2).⁵

Figure 2. DG025 transition state analogue in complex with IRAP (PDB: 4Z7I). The Electrostatic Complementarity™ (EC) isosurface has been rendered over the active site of the protein. All images were produced in Flare™. ^3,6

Natively, DG025 is a pseudo-decapeptide, which, for this purpose, was truncated to the first four residues. This query was submitted to Blaze and a virtual screening experiment performed over all commercial databases in addition to ChEMBL. All hits were screened to remove unwanted chemistry using the PAINS filters and finally clustered by chemical similarity. From this, the top scoring hit in each cluster was purchased and tested in vitro.

Blaze identified hit compound 1 (Figure 3), which displayed modest potency (IC₅₀ = 55 μM). The first, and most important thing to notice, is that Blaze retrieved a small molecule, despite starting from a pseudopeptide query. This is an exemplification of the benefits of virtual screening in 3D field descriptor space, as field points are ‘chemistry agnostic’, opening the potential to retrieve compounds with low 2D similarity but high 3D field similarity. As can be shown in Figure 3, the field overlays of compound 1 and the quadramer pseudopeptide query present with clustered electrostatically negative and hydrophobic field points over the amide, pyrazole and azepane regions of the scaffold respectively. Understandably, in the light of these conserved field motifs, Blaze retrieved a compound presenting with the 3D electrostatic and shape features required to bind the same region of the IRAP target site, in the same binding conformation, as the quadramer pseudopeptide query.

Figure 3. Blaze query compound (left), Blaze hit compound 1 (right) and the 3D overlay of hit compound 1 and the Blaze query (centre). Red field points are electrostatically positive, blue points are electrostatically negative, yellow are shape and orange are hydrophobic.

Notwithstanding, it is important to note that this work simply did not finish with the identification of compound 1 from the virtual screen. Rather, the authors used compound 1 as an initial starting point for a medicinal chemistry optimization campaign to improve not only the potency, but also the physicochemical properties of compound 1. Explicitly, the terminal phenyl and azepane substituents were iteratively transformed, through an intermediate compound 2, to give compound 3 (Figure 4). Further in vitro testing showed that compound 3 is a highly selective IRAP inhibitor, binding the catalytically active open conformation of IRAP with an IC₅₀ = 157 nM.

Figure 4. Molecular structures of initial hit compound 1, intermediately optimized compound 2 and resulting compound 3, after further optimizations. The circles over compound 1 highlight the focus of the medicinal chemistry program.

Rapidly progressing molecule discovery projects using Blaze

By virtual screening in 3D field descriptor space, the authors were able to use a peptide hit to search small molecule chemical space to identify hits that could start a small molecule medicinal chemistry program against the target. The hit to lead derivatives achieved greater potency and the desired physicochemical properties. This exemplifies the power of Blaze to identify new, yet relevant, chemical matter required to progress molecule design projects towards potent and promising compounds.

References

1. Mpakali, A.; Georgaki, G.; Buson, A.; Findlay, A. D.; Foot, J. S.; Mauvais, F.; van Endert, P.; Giastas, P.; Hamprecht, D. W.; Stratikos, E. Stabilization of the Open Conformation Οf Insulin‐regulated Aminopeptidase by a Novel Substrate‐selective Small‐molecule Inhibitor. Protein Sci 2024, 33 (9), e5151. https://doi.org/10.1002/pro.5151
2. Cheeseright, T. J.; Mackey, M. D.; Melville, J. L.; Vinter, J. G. FieldScreen: Virtual Screening Using Molecular Fields. Application to the DUD Data Set. J. Chem. Inf. Model. 2008, 48 (11), 2108–2117. https://doi.org/10.1021/ci800110p
3. Cheeseright, T.; Mackey, M.; Rose, S.; Vinter, A. Molecular Field Extrema as Descriptors of Biological Activity:  Definition and Validation. J. Chem. Inf. Model. 2006, 46 (2), 665–676. https://doi.org/10.1021/ci050357s
4. Grant, J. A.; Gallardo, M. A.; Pickup, B. T. A Fast Method of Molecular Shape Comparison: A Simple Application of a Gaussian Description of Molecular Shape. J. Comput. Chem. 1996, 17 (14), 1653–1666. https://doi.org/10.1002/(SICI)1096-987X(19961115)17:14<1653::AID-JCC7>3.0.CO;2-K
5. Mpakali, A.; Saridakis, E.; Harlos, K.; Zhao, Y.; Papakyriakou, A.; Kokkala, P.; Georgiadis, D.; Stratikos, E. Crystal Structure of Insulin-Regulated Aminopeptidase with Bound Substrate Analogue Provides Insight on Antigenic Epitope Precursor Recognition and Processing. J. Immunol. 2015, 195 (6), 2842–2851. https://doi.org/10.4049/jimmunol.1501103
6. Bauer, M. R.; Mackey, M. D. Electrostatic Complementarity as a Fast and Effective Tool to Optimize Binding and Selectivity of Protein-Ligand Complexes. J. Med. Chem. 2019, 62 (6), 3036–3050. https://doi.org/10.1021/acs.jmedchem.8b01925