Crystallographic fragment based lead discovery is a now a routine technique, which can sample 1000’s of compounds per week. But how do we identify the most appropriate compounds to screen against our target of interest?

Crystallographic fragment based lead discovery

Briefly, small fragment molecules (<300 Da) explore chemical space more efficiently as at their smaller size, as order of magnitudes fewer possible compounds. The simpler moieties in fragments also allow detection of smaller interaction surfaces, and hopefully more specific interactions, within a protein. Crystallography allows identification of the binding pose and atomic interactions, but gives no affinity information. Affinity information for fragments can be difficult to obtain due to their small size typically leading to low affinity binding. Developing a fragment into a lead compound involves many steps of identifying possible interactions, sythensising or purchasing new compounds to explore those hypothesised reactions and then soaking or co-crystallsing to obtain the complex structure. Furthermore, an biophysical screen or assay to determine affinity and efficacy should also be designed to rank the potential lead candidates.

Choosing a Fragment Library

XChem, the crystallographic screening facility at Diamond Light Source maintains access to multiple screening libraries for it’s users, but historically the first screening on most projects has been done with the Diamond Science Poised Library (DSPL/ DSiP) [1]. This library is focused on synthetic accessibility, with molecules being formed from two pieces, or synthons which are joined by common medicinal chemistry reactions. This allows for replacement of chemical moieties for quick synthesis or purchase. This library has often been suggested as a first library, as the XChem team have links with industrial and academic production of follow up compounds. Furthermore, these compounds form the part of software efforts (Fragalysis) which show vectors on which the molecule can be expanded, using molecules found in the Enamine REAL database. This relies on a graph network [2] to explore synthetically accessible chemical space, which can be adapted to include a wider variety of starting molecules and synthetic reactions.

However, future synthetic accessibility of further compounds is only useful if you have learnt a sufficient amount about your protein binding site in the first screening campaigns. Alternatively, screening smaller libraries with smaller, but representative compounds may tell the lead scientist more about the exploitable interactions on the protein. However, such smaller compounds are unlikely to be immediately developable into higher potency follow ups.

Fraglites are a a set of 31 halogenated small fragments, designed to interaction patterns in the binding site [3]. They are designed to have two hydrogen binding sites separated by between 1 and 5 bonds. The halogen atom is used for detection of the fragment in anomalous difference density maps. Due to their small size, they are unlikely to be detected at normal soaking concentrations without such a moiety. The detection by anomalous difference density will likely need changes to the XChem workflow for hit detection, if these are to be used as an initial screen.

MiniFrags are a set of 81 low molecular weight (5-7 heavy atom) fragments which are stable in aqueous solution at high (1M) concentrations [4]. Astex’s intended use of this fragment set is to discover a wider number of sites on the protein, with a high hit rate, to identify alternate interaction to grow/ merge existing fragment compounds into. The high soaking concentration is unlikely to be amenable to all crystal systems. It will need care integrating in a user program as the compounds need individual preparation rather than by a prepared plate of compounds for soaking, that can be dispensed using acoustic droplet ejection.

Summary of Library properties: MiniFrag & DSiP & Fraglites

Using these small libraries at the first stage of crystal system characterisation should lead to high initial hit rates, and site discovery. On the XChem platform, this would be done during a pre-screen which is typically 100 crystals. Alongside work to priortise warm and hot spots [5] in proteins, we aim to see whether these small fragments provide sufficient benefit to finding new sites, and developing fragments to highly ligand efficient potency faster.

References

[1] Cox, O. B., & et al. (2016). A poised fragment library enables rapid synthetic expansion yielding the first reported inhibitors of PHIP(2), an atypical bromodomain. Chem. Sci., 7(3), 2322–2330. https://doi.org/10.1039/C5SC03115J

[2] Hall, R. J., Murray, C. W., & Verdonk, M. L. (2017). The Fragment Network: A Chemistry Recommendation Engine Built Using a Graph Database. Journal of Medicinal Chemistry, 60(14), 6440–6450. https://doi.org/10.1021/acs.jmedchem.7b00809

[3] Wood, D. J., Lopez-Fernandez, J. D., Knight, L. E., Al-Khawaldeh, I., Gai, C., Lin, S., … Waring, M. J. (2019). FragLites – Minimal, Halogenated Fragments Displaying Pharmacophore Doublets. An Efficient Approach to Druggability Assessment and Hit Generation. Journal of Medicinal Chemistry, 62(7), 3741–3752. https://doi.org/10.1021/acs.jmedchem.9b00304

[4] O’Reilly, M., Cleasby, A., Davies, T. G., Hall, R. J., Ludlow, R. F., Murray, C. W., … Jhoti, H. (2019). Crystallographic screening using ultra-low-molecular-weight ligands to guide drug design. Drug Discovery Today, 24(5), 1081–1086. https://doi.org/10.1016/j.drudis.2019.03.009

[5] Rathi, P. C., Ludlow, R. F., Hall, R. J., Murray, C. W., Mortenson, P. N., & Verdonk, M. L. (2017). Predicting “Hot” and “Warm” Spots for Fragment Binding. Journal of Medicinal Chemistry, 60(9), 4036–4046. https://doi.org/10.1021/acs.jmedchem.7b00366