Construction of the CSIRO Fragment Library Craig L. Francis, A Peter W. Kenny, B Olan Dolezal, B Simon Saubern, A Megan Kruger, A G. Paul Savage, A Thomas S. Peat, B,C and John H. Ryan A,C A CSIRO Materials Science and Engineering, Bayview Avenue, Clayton, Vic. 3168, Australia. B CSIRO Materials Science and Engineering, 343 Royal Parade, Parkville, Vic. 3052, Australia. C Corresponding authors. Email: tom.peat@csiro.au; jack.ryan@csiro.au A fundamental component of a successful fragment screening program is a productive fragment library, one that delivers hit fragments with potential for pharmaceutical development. A proprietary fragment library was developed by identifying and extracting subsets of CSIRO’s Compound Collection using two complimentary approaches. Over time, the use of surface plasmon resonance as a front-line screening tool has enabled identification and exclusion of problematic compounds and led to a more reliable fragment screening library. Manuscript received: 25 June 2013. Manuscript accepted: 12 September 2013. Published online: 17 October 2013. Introduction In 2011, Zelboraf (vemurafenib) was approved by the US Food and Drug Administration (FDA) for the treatment of late-stage melanoma. [1] This approval was a milestone for the field of fragment-based drug discovery (FBDD), as it is considered the first approval of a drug derived from a fragment hit, albeit one discovered through a high-throughput screening campaign. [2,3] As of 2011, there were 22 compounds from FBDD in clinical trials across a range of indications such as cancer (liver, leukaemia, myeloma, gastrointestinal stromal tumour, breast), coronary artery disease, antibacterial, diabetes, pain, chronic obstructive pulmonary disease (COPD), and psoriasis, indi- cating the utility of the fragment-based approach. [1] There are three key stages to the fragment-based approach: fragment library design, fragment library screening, and fragment elab- oration. [4] Herein, we focus on the design of a fragment library and integrating this library with screening methods. Typically, molecules need to meet a range of criteria in order to be considered for inclusion into a fragment library [5–7] and there is recognition of the need to tailor the choice of fragments depending on the screening method, e.g. by NMR, X-ray crystallography, or biochemical assay. [8–10] Molecules are included based on a modified Lipinski Rule of 5, [11,12] referred to as the Rule of 3: a set of physiochemical criteria including primarily molecular weight (MW) , 300, number of H-bond donors (HBD) # 3, number of H-bond acceptors (HBA) # 3, and cLogP # 3, but also considering the number of rotatable bonds # 3 and polar surface area # 60 A ˚ 2 . [13] A lower limit to the molecular weight of 100–150 is commonly employed, [7] which minimizes the chance of fragments that can confound lead generation owing to binding to multiple sites of a target. [14] It is considered advantageous to include molecules with func- tionalities that facilitate future fragment elaboration while avoiding reactive, unstable or toxic scaffolds such as alkylating and acylating groups. [15] Additionally, it is recommended that functional groups or substructures known to be associated with toxic side-effects such as carcinogenicity or with metabolic liabilities be excluded. [16] As some marketed drugs contain problematic functionalities such as a nitro group, it has been suggested that fragments with such functionalities could be included with a view to bioisostere replacement at a later stage. [4,17] Fragment libraries are available commercially; how- ever, it has been noted that commercial libraries can be limited in terms of chemotype and shape diversity. [4] Alternatively, libraries can be constructed by selecting individual molecules from commercial vendors or from proprietary collections. In so doing, consideration is often given to maximizing the diversity of a given library in terms of representative ring systems and/or scaffolds, side chains, and overall shape. [12] The fragment libraries have to be tailored for fragment- screening methods. Given that the binding affinities of fragment hits are typically in the range of 0.2–10 mM, sensitive biophysical techniques are typically required to detect binding, and thus fragments require good aqueous solubility, typically at least 1.0 mM in aqueous buffer. [12] Although the solubility of a given fragment can be determined experimentally, the prediction of water solubility is still a challenge. [18] In order to build our fragment-based screening efforts, [19–22] we undertook several approaches to constructing a fragment library. To initiate the program, we purchased a commercial fragment library. We followed that with selection of fragment- like molecules from CSIRO’s proprietary compound collection based on first selecting representative examples of heterocycles within the collection, and second in silico techniques for select- ing a set of fragment-like molecules that represent the diversity of the collection. Herein, we describe these approaches and CSIRO PUBLISHING Aust. J. Chem. 2013, 66, 1473–1482 http://dx.doi.org/10.1071/CH13325 Journal compilation Ó CSIRO 2013 www.publish.csiro.au/journals/ajc Full Paper RESEARCH FRONT