A Collaborative Web-based Architecture for! Fragment-Based Drug Discovery Data Michael Siani-Rose1, Barry Bunin, Ph.D.1, Norah MacCuish, Ph.D.2 1Collaborative Drug Discovery, 1633 Bay Shore Hwy, Suite 342, Burlingame, CA 94010 2Mesa Analytics & Computing, Inc. 212 Corona Street, Santa Fe, NM 87501

Abstract! Hsp90 fragment with assay data Analysis of Fragments and ! Screening against ChemBridge The Collaborative Drug Discovery (CDD) Vault™ for private Bound Ligand Crystal Structures Fragment-Based Drug Discovery (FBDD) data also contains Astex RO3 Fragment Library Fragment-Based Public data (such as the ChemBridge “Astex Rule of 3” compliant Fragment Library). The CDD platform provides a secure, cloud-based database with a web interface that permits users to access and search chemical FBDD collections of structures with SAR data, as well as screening and preclinical study data.

An open call is being made to all FBDD researchers, for input into a general purpose resource of not only fragments, but also associated, published or patented bioactivity data associated with published fragments. It is envisioned that this will be a unique data source to answer common questions in the field, such as why certain fragments bind better than others, chemical property trends, 2D vs. 3D fragment properties, and other useful similar insights that can only be addressed with a critical mass of data. The Public resource would complement researchersʼ private mining of IP-sensitive data in the secure Vaults. Publications, patents, or data may be provided for consideration, formatting and free hosting with attribution for the public, scientific good conveniently via email to Figure 3. Example Shape-Pharmacophore Analysis. 2D Hsp90 Figure 4. Results from a similarity search in the CDD Vault for an [email protected]. ligands structures were downloaded from the CDD Vault. Crystal active query (shown in the top left) against the ChemBridge structures (PDB 2012, Murray 2010) were utilized for 3D shape- Fragment Database; results (including structure and Tanimoto pharmacophore clustering, showing additional binding complexity over similarity measures) are shown in blue on the right side of the figure. Introduction! the 2D series in Murray, et al. (Murray 2010). Methods Figure 2. Hsp90 fragments loaded into CDD Project (Murray 2010).! We have built a database of fragments from the literature for drug discovery applied to the chaperone Hsp90, which plays a role in conformational stability, maturation and functionality of proteins in the Collaboration in FBDD! cell. Inhibitors have been developed against Hsp90 as To date, drug discovery pharma and biotech companies have maintained chemotherapeutic agents in cancer. Murray et al. (Murray 2010) found Figure 1. Fragment-based drug discovery. Fragments are used to proprietary libraries of compounds, cautiously guarding their methods for two leads with high Ligand Efficiency (LE) which bind in different gradually fill the site through growth (top) or linking of adjacent fragments building the libraries, which give them an advantage based on techniques configurations against the Hsp90 active site using fragment based (bottom), then linked or augmented further to build tighter-binding ligands. of fragment-based diversity as well as their preferred medicinal chemistry. discovery. We built a database within the CDD database environment, Academic labs, while open about their approaches, individually lack the capturing 2D-chemical structure, ITC (isothermal titration calorimetry) Fragment-based drug discovery is complementary to traditional drug resources to build and refine (over years and decades) large optimized and IC50 (uM) data. This database facilitates analysis of Murray et discovery methods, (1) utilizing intelligently designed libraries which libraries for screening. al.ʼs two lead series against Hsp90: aminopyrimidines and phenols. ! obviate the need for High Throughput Screening (HTS) of 103-106 compounds, (2) using a more flexible approach to expanding a drug lead A repository for fragments and binding for lead or drug development data than either human- or algorithm-driven medicinal chemistry. And (3) would provide a tremendous resource for cross-pollination of fragment- References! making use of actual protein structure techniques (e.g., X-ray based discovery, which is often hidden behind closed doors today. As with 1. Balloon v 1.3.1.983, URL users.abo.fi/mivainio/balloon/index.php. 2. ChemBridge Fragment Library URL www.chembridge.com/screening_libraries/ crystallography, NMR-assisted or a variant), building up a better drug in all drug discovery efforts (including HTS of large compound libraries) fragment_library/ Figure 5. Example Shape-Pharmacophore Similarity Search. A the context of the protein binding site. published data is often limited to positive results. Negative data can 3. Erlanson DA. Fragment-based lead discovery: a chemical update. Curr Opin Biotechnol. greatly augment our understanding of what a good drug should look like 2006 17(6):643-52. single ligand crystal structure was used as a query (upper left image) 4. Fingerprint Module v. 2.0, Mesa Analytics & Computing, Inc., URL www.mesaac.com/ against 26 thousand conformations generated from the ChemBridge Converting a fragment to a lead molecule can be done using the following by allowing us to avoid certain pitfalls. site_media/uploads/files/FingerprintModule2_0.html (2012). Fragment library, finding four similar conformations of four different methods: 5. Grouping Module v. 2.0, Mesa Analytics & Computing, Inc., URL www.mesaac.com/ structures. Image on lower right shows ChemBridge structures 1. Fragment optimization through extending molecules by standard Several issues can act as barriers to effective cross-laboratory site_media/uploads/files/GroupingModule2_0.html (2012). 6. Jmol: an open-source Java viewer for chemical structures in 3D. http://www.jmol.org/ aligned to crystal structure. medicinal chemistry techniques. collaborations; heterogeneity of data (wide range of variability observed for 7. Murray CW et al. Fragment-based drug discovery applied to Hsp90. Discovery of two lead 2. Merging and linking of fragments by combining multiple fragments, in vitro IC50 calculations between laboratories, lot to lot differences in series with high ligand efficiency. J Med Chem. 2010 53(16):5942-55. which have been shown to bind separately near each other. compounds tested which both require normalization or monitoring) and 8. O'Boyle NM, Banck M, James CA, Morley C, Vandermeersch T, Hutchison GR, "Open Babel: An open chemical toolbox." J. Cheminf. 2011, 3, 33. DOI:10.1186/1758-2946-3-33. Acknowledgments! intellectual property (IP) ownership. Data obtained in different laboratories 9. The Open Babel Package, version 2.3.1 openbabel.org. The authors would like to thank Dan Erlanson, Carmot, Inc. for are often stored in incompatible formats and site-licensed databases, 10. PDB, URL http://www.rcsb.org/pdb/ (2012). productive conversations and Open Source Software packages: 11. R Development Core Team (2008). R: A language and environment for statistical For Further Information! preventing effective data mining across labs, especially for accessing Balloon (Balloon 2012), OpenBabel, R (R 2008), Jmol (Jmol 2012). negative data. In addition, data persistence is uncertain when employees computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, Please contract [email protected] URL www.R-project.org. and students leave or projects wind down, resulting in data often 12. ShapeBase Module 2.0, Mesa Analytics & Computing, Inc., URL www.mesaac.com (2012) remaining fragmented and inaccessible. 13. Woodhead AJ et al. Discovery of (2,4-dihydroxy-5-isopropylphenyl)-[5-(4-methylpiperazin-1- ylmethyl)-1,3-dihydrois oindol-2-yl]methanone (AT13387), a novel inhibitor of the molecular chaperone Hsp90 by fragment based drug design. J Med Chem. 2010 53(16):5956-69.