Challenges of fragment screening

Fragment-based screening for lead generation has seen tremendous growth and success in the last few years. Furthermore, the careful design of fragment libraries has ensured both that the coverage of chemical space is as great as possible and also that the included fragments have desirable physical properties. This particular effort has thus enhanced the advantages of a fragment-based approach. Other technological advances and applications have increased the speed of the process, resulting in more successful case studies being presented. Nonetheless challenges still remain. Weak fragment hits are often overlooked in favor of more potent HTS hits that may have poorer physical properties and ligand efficiencies. The timelines necessary to realize success with fragment-based screens can be long relative to other lead generation approaches, since fragment hits need to be given sufficient consideration and may require more cycles for optimization. A robust system for crystallography, which can be difficult to develop, can also dramatically affect the final outcome of a fragment-based lead generation campaign by enabling the determination of high-resolution complex structures that can serve as starting points for structure-based design. Active research is helping to address these challenges. Computational approaches that can aid in the optimization process either through growing or linking of fragments continue to be developed and can play a significant role in reducing the time required for improving the potency of an initial fragment lead. Methods aimed at exploiting fragment hits for uses other than scaffold generation are also being established to take full advantage of this information as well as any associated structural information available for a project. For example, fragment hits can be merged onto an HTS scaffold during the lead optimization process. In addition, for a given target, potential pharmacophores can be derived from fragment hits and later used for virtual screening of databases to enable scaffold hopping. Fragment positioning methods such as GRID [1], MCSS [2–4], SPROUT [5], MUSIC [6], LUDI [7, 8], and Superstar [9] have been in use for over two decades now and are typically employed during the early stage of lead optimization. These methods determine energetically favorable binding site positions for various functional group types or chemical fragments based on molecular mechanics or knowledge-based potentials. ‘‘Hot spots’’ can be calculated for a wide range of functional groups in a given target binding site/region. Such target-derived pharmacophoric points can also be used to guide docking calculations to more finely sample the relevant regions of a binding site (e.g., [10, 11]) or to perform pharmacophore searches of large databases. Caveat [12] and HOOK [13] were among the first ‘‘fragment-linking’’ computational approaches developed in the early 1990s. Newer generation computational methods continue to be developed (e.g., Re-core [14], Allegrow (Boston De Novo Design, Boston, MA, 2009), Confirm [15], MED-SuMo [16], and pharmacophore modeling for scaffold replacement in MOE (Chemical Computing Group, Montreal, Canada, 2009)) and successful uses of these newer programs are being reported. Increasingly, standard molecular docking programs are being utilized to screen large databases of small molecules to created target-focused fragment sets for experimental testing either in a high-concentration biochemical assay or by biophysical means such as NMR, Biacore, mass spectrometry, or X-ray crystallography. Focused fragment sets D. Joseph-McCarthy (&) Infection Computational Sciences, AstraZeneca R&D Boston, 35 Gatehouse Drive, Waltham, MA 02451, USA e-mail: [email protected]