1. The Dengue Docking Project: Virtual Screening on the ProtoGRID
EGEE'06 conference, Swiss Grid Track
Dr. Michael Podvinec
Biozentrum, University of Basel &
Swiss Institute of Bioinformatics
Basel, Switzerland.

3. Areas infested with Aedes aegypti
Areas with Aedes aegypti and recent epidemic dengue
Data: CDC, 1999

4. Every year: 2.5 billion are at risk of infection
million people get infected with Dengue
500'000 cases of Dengue Hemorrhagic Fever
25'000 deaths
No vaccination.
No specific treatment available.

5. Public-private partnership
Biozentrum, University of Basel : in silico docking
NITD: In vitro/in vivo follow-up
Novartis: Drug development at cost
Schrödinger, Llc.: Scientific collaboration on docking
SwissBioGrid Grid computing resources, Proof of Concept for SwissBioGrid

6. What does it take to make a drug?
Target validation
Opti- mization
Target ID
Screening
Preclinical
Clinical
HIT
LEAD
CANDIDATE DRUG
DRUG
~12 years
802 mio US$ (DiMasi, J.A. et al. (2003) J Health Econ, 22, ).
only 1 in 10‘000 NCE survives (Heilman, R.D. (1995) Qual Assur 4(1) 75-9.)

7. Virtual screening to the rescue?
Target validation
Opti- mization
Target ID
Screening
Preclinical
Clinical
HIT
LEAD
CANDIDATE DRUG
DRUG
 Virtual screening (computer simulations)
Save initial investment (HTS)
Predict likely hits in silico
GRID-based
Still must assay results

8. availability of compounds (or synthetic accessibilty)
Chemical space
10E29 derivatives of n-Hexane Need some way to find an interesting region of chemical space
availability of compounds (or synthetic accessibilty)
natural compounds
similarity to other drugs
shape/feature complementarity with known drugs/inhibitors
Camille Pissarro, Haystack, 1873

9. Protein-ligand complexes share common characteristics
Protein-ligand complexes share common characteristics. We can use these characteristics to predict ligands for target molecules.

10. Structure and serotype analysis of dengue target sites:
1: Envelope Glycoprotein E
b-OG
catalytic triad
P1Arg contact
2: NS3 serine protease
S-Adenosyl- homocysteine
GMP analogue
3: NS3 RNA helicase
4: NS5 RNA methyl transferase

11. SwissBioGrid Initiative
Federate resources using GRID technology
provide computing platform
enable Dengue proof-of-concept
[BC]2 Linux cluster
[BC]2 UD PC grid
Uni ZH AIX cluster
Basel
Zürich
VitalIT Linux cluster
CERN
International
collaborations
CSCS UNIX HPC resources
EGEE
Lausanne
Lugano

12. Phase I (protoGRID) Phase II (NorduGRID)
Evaluate requirements for productive GRID infrastructure
Federate pre-existing Swiss HPC ressources (non-intrusive system - cycle stealing ONLY)
Swiss HPC ressources are heterogeneous: Clusters, HPC, PC Desktop Grids
Geographically and administratively separate entities
Support the specific data requirements of bioinformatics applications
Validation of docking results on heterogeneous architectures
Productive services for the scientific PoC (Docking against Dengue targets)
Phase II (NorduGRID)
Extend NorduGRID based on lessons learnt from ProtoGRID (data model, LRMS support)
Productive services for several scientific projects (e.g. peptide mass fingerprinting in proteomics)

13. SBG Phase 1: ProtoGRID Single interface to Swiss HPC resourcesQW QW
CSCS (PBS, Itanium 64) Ticino
[BC]2 PC Desktop Grid (UD MP, Win32) Basel
Grid Node Manager
Grid Data Manager QW QW
[BC]2 HPC cluster (SGE, x86-32) Basel
Vital-IT HPC cluster (LSF, Itanium 64/ Nocona) Lausanne

14. Some hurdles in grid adoption [as identified in the SwissBioGrid development process]
Where do the CPUs come from?
Most HPC resources are busy already
Agree on dedicated compute time for grid projects?
Buy new clusters for your grid?
PC Desktop grids provide a huge untapped resource
Licensing schemes of most commercial software are not suitable for GRIDs
Application clearing:
Ensuring data integrity on distributed resources
Non-intrusiveness of the application
Security issues: Avoid accidental or malicious negative impact on running systems
Numerical stability in heterogeneous environments
Data model in bioinformatics is different from HEP
Most applications need access to large public databases

15. Achievements of GRID-enabled Dengue Docking
Technical: Phase I SwissBioGrid infrastructure complete
large-scale parameterization test using Autodock 3.0.5: 500‘000 docking runs, 38‘000h CPU time
Prioritized hit list from 127k library screening (GLIDE) is undergoing In vitro testing at NITD
Some initial hits (at 20 M)
Currently screening the ZINC open source compound library (~4 Mio cpd)

16. Acknowledgements Biozentrum & SIB CSCS Novartis Institute for
University of Basel:
Torsten Schwede
Jürgen Kopp
Marco Scarsi
Rainer Pöhlmann
Konstantin Arnold
CSCS
Marie-Christine Sawley
Peter Kunszt
Sergio Maffioletti
Novartis Institute for
Tropical Diseases
Alex Matter
Mark Schreiber
Subhash Vasudevan
Siew Pheng Lim
Fraunhofer SCAI:
Martin Hoffmann
Marc Zimmermann
Vital-IT:
Victor Jongerneel
Bruno Nyffeler
Jacques Rougemont
Heinz Stockinger
Arthur Thomas
Novartis
Manuel Peitsch
René Ziegler
Eric Vangrevelinghe
Pascal Afflard
EGEE/CNRS:
Vincent Breton
Nicolas Jacq
Schrödinger, Inc:
Jörg Weiser