Fragment-based drug discovery (FBDD) describes the creation of a drug compound via building up from small, weakly binding molecules and successive modifications to improve potency and ligand efficiency. Small molecular weight fragments that bind weakly, but form high quality interactions with a protein target, are selected to optimise into larger, more potent molecules(Jencks, 1981, Scott et al., 2012). Initial fragment molecules tend to conform to the Rule of Three(Congreve et al., 2003), a standard rule of thumb for determining optimal fragment ligand properties: a
molecular weight of less than 300 Da, a calculated logP of ≤ 3, three or fewer
hydrogen bond donors, and up to three hydrogen bond acceptors.
Using an appropriate fragment library, compounds are screened using one of several biophysical techniques to detect weak non-covalent interactions, after which fragment
‘elaboration’ occurs in which validated hits undergo cycles of synthesis into larger
compounds with input from structural biology, medicinal chemistry and computational chemistry. This eventually produces a potent compound.
1.4.1 Other Techniques in FBDD
Several biophysical techniques are used by FBDD users during the early stages of development. X-ray crystallography is the generally considered to be the most powerful primary screening technique by FBDD practitioners. This generates three- dimensional structures of protein-ligand complexes at atomic resolution. These structures are considered very important for validating hits, as well as for establishing
initial binding modes. However this is dependent on access to synchrotrons, as well as high quality crystals, which may not always be possible.
Native mass spectrometry (MS) is extremely versatile. Protein/fragment mixtures undergo electrospray ionization (ESI) and fragment binding can be observed as a corresponding increase in the mass of the target. This way, fragments can be screened in large cocktails, and a gauge on affinity gained from the relative abundance of different protein-ligand species(Vivat Hannah et al., 2010). However the requirement for relatively large amount of target limits the utility of this technique.
Another technique is surface plasmon resonance (SPR), in which the protein target is covalently bound to the gold surface of an SPR chip, and solutions of individual ligands are then passed over it. If a fragment binds to the target, an increase in mass is detected, and from the resulting association/dissociation curve the binding kinetics and affinity can be calculated(Navratilova and Hopkins, 2010). This provides information for kon and koff, rather than simply KD and so might be more suited for
follow up studies rather than initial screening.
Typically a range of techniques is employed in order to ensure results are validated. There is a distinct lack of correlation between fragment hits obtained via different techniques, in fact it is possible to run a fragment screen using two different methods on an identical library and arrive at a dramatically different set of hits(Wielens et al., 2013).
1.4.2 Other NMR Techniques in FBDD
The principal NMR method employed in FBDD – other than ligand-observed experiments – is chemical shift perturbation mapping (CSP). Here, two 2D Heteronuclear single quantum coherence spectroscopy (HSQC) experiments are run in the absence and presence of a ligand. In a 15N HSQC spectrum each peak is representative of an amide proton, thus representing a particular amino acid. Any shift of a particular amide proton upon ligand binding is indicative of ligand binding. In contrast to ligand-observed NMR, this is very much protein-observed. The method relies on chemical shifts of amide peaks of the protein target being acutely sensitive to changes in local environment.
It also depends upon isotopic enrichment of protein (15N) since the natural abundance
of this spin ½, NMR-active nucleus, is only 0.368%. This process can be tricky and costly, and is a clear limitation.
CSPs can be used as an initial screen on a library of ligands in order to identify binders but is more likely to be employed as a secondary method in order to give more information. Both the interface and the kinetics of binding can be identified by titration of increasing quantities of ligand(Medek et al., 2000). Given fast exchange between protein and ligand, incrementally increasing the ligand concentration produces a trajectory of CSPs for certain amide peaks, these can then be fitted to determine the dissociation constant(Williamson, 2013).
1.4.3 The fruits of FBDD
In 2011 a phase 3 randomised clinical trial of 675 patients with untreated metastatic melanoma taking the drug vemurafenib – who possessed the BRAF V600E mutation
– showed improved rates of overall survival (OS) and progression-free survival (PFS)
over the previous standard therapy, dacarbazine(Chapman et al., 2011). Later that year the FDA approved the drug, and it became the first drug to be approved that had been produced with fragment-based principles.
The discovery of vemurafenib began with an initial screen of 20,000 compounds between 150 and 350 Daltons binding to various kinases by in vitro phosphorylation measurement. Of these, 238 compounds were found to bind to three kinases and subsequently > 100 bound crystal structures were solved(Tsai et al., 2008). Using a structure-guided approach the potent, selective inhibitor was subsequently found to inhibit BRAF V600E with an IC50 of 13 nM.
Whilst this significant milestone for FBDD was passed in 2011, the future holds the prospect of much greater reward. In phase 3 trials currently is the BACE inhibitor MK-8931 and a trial involving 1500 patients with Alzheimer’s disease set to be completed in 2018. In phase 2 trials are many FBDD-derived compounds for a variety of disease indications (including multiple myeloma, non-Hodgkin’s lymphoma, non- small cell lung cancer, and gastrointestinal stromal tumour), among them are compounds that inhibit: CDKs 1, 2, 4 and 5(Wyatt et al., 2008), VEGF(Albert et al., 2006), JAK2(Howard et al., 2009) and Hsp90(Murray et al., 2010, Woodhead et al., 2010).