Grid computing is currently developing into a major driving force for new approaches towards collaborative large scale science. Several national and international eScience programs have fostered collaboration between researchers from different scientific domains and the first large scale experiments on grids infrastructures such as EGEE are underway.

In the biomedical community, grid computing has initiated several projects on large scale in silico drug screening approaches. The project WISDOM was amongst the first projects in the public domain that made use of grid enabled in silico docking to simulate the interaction of potential drugs with target proteins. In silico docking is the first step in the virtual screening process, which is one of the most promising approaches to speed-up and to reduce the costs of the development of new drugs. WISDOM was organised in the course of the first biomedical data challenge for drug discovery on the EGEE grid production service and was effectively run between 11 July 2005 and 31 August 2005. In the course of this first real large scale biomedical application on EGEE, we have executed more than 46 million individual docking experiments.


A neglected disease has been selected as target to prove the feasibility of the concept: malaria. The number of cases and deaths from malaria increases in many parts of the world. There are 300 to 500 million new infections, 1 to 3 million new deaths and a 1 to 4% loss of gross domestic product (at least $12 billion) annually in Africa caused by malaria. For both vector control and chemotherapy, knowing the gene sequences of Anopheles and Plasmodium species should lead to discovery of targets against which new insecticides or anti-malarial drugs can be produced. However, such discoveries are likely to be patented and only developed at prices unaffordable to governments or villagers in tropical countries.

Malaria Parasites

Malaria parasites are micro-organisms that belong to the genus Plasmodium. There are more than 100 species of Plasmodium, which can infect many animal species such as reptiles, birds, and various mammals. Only four species of Plasmodium infect humans in nature.

Plasmodium falciparum, which is found worldwide in tropical and subtropical areas, is the only species that can cause severe, potentially fatal malaria. It is estimated that every year 700,000 to 2.7 million people are killed by P. falciparum, especially in Africa where this species predominates.

The Experiment

As ligands we used a subset of the ZINC database (http://zinc.docking.org), a free database of chemical compounds in a ready-to-dock format. 1,000,000 compounds were selected: the ChemBridge database (~500,000) and 500,000 drug like compounds. They were docked against 5 different structures of plasmepsin, a protein target from the Malaria parasite Plasmodium falciparum. This molecule is potentially very interesting because it acts on human haemoglobin. Inhibiting its action would prevent the parasite from feeding itself with the human blood. We used three Plasmepsin II structures (1lee, 1lf2, 1lf3) and one Plasmepsin IV structure (1ls5) from the Brookhaven protein database (PDB, www.rcsb.org/pdb) as targets. Two widely used docking algorithms were applied to find the best hits. AutoDock (v3.05) (http://w3.to/autodock) is an open source algorithm developed by the Scripps Research Institute. 

Fraunhofer SCAI is known world-wide for developing one of the best docking algorithms, FlexX (v2.0) (http://www.biosolveit.de/). SCAI gave access to FlexX within the framework of this collaboration for a limited time. A server solution was studied to solve the license issue. We have prepared 8 target scenarios for FlexX and 10 different target scenarios for AutoDock based on the inclusion of water molecules in the proteins during the docking process. Additionally, we have created 4 different parameter sets for FlexX based on the variations in the software parameters and two different parameter sets for AutoDock.

Docking tools


AutoDock is a suite of automated docking tools. It is designed to predict how small molecules, such as substrates or drug candidates, bind to a receptor of known 3D structure. AutoDock has applications in:

  • X-ray crystallography
  • structure-based drug design
  • lead optimization
  • virtual screening (HTS)
  • combinatorial library design
  • protein-protein docking
  • chemical mechanism studies
  • http://autodock.scripps.edu/

Morris, G. M.; Goodsell, D. S.; Huey, R.; Hart, W. E.; Halliday, R. S.; Belew, R. K.; Olson, A. J. Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J Comp Chem 1998, 19, 1639-1662.


FlexX is an extremely fast, robust, and highly configurable (FlexX-able) computer program for predicting protein-ligand interactions. Its main applications are:

  • Binding mode prediction
  • Virtual high-throughput screening (vHTS)

Rarey, M.; Kramer, B.; Lengauer, T.; Klebe, G. A fast flexible docking method using an incremental construction algorithm. J Mol Biol 1996, 261, 470–489.


The goal in result analysis was to identify novel antimalarial compounds and to compare the used docking tools, parameters and targets and to draw some general conclusions about their impact on virtual screening results.

General Statistics

The different parameter settings and target scenarios had severe influence on the virtual screening results. Although there could be found a high correlation between the scores of the docked compounds under different conditions the differences were significant and partly different compounds were selected as the top-scoring ligands. But the highest diversity was produced by the application of the different docking tools. There was almost no dependence between the docking scores calculated by AutoDock and FlexX. When looking at the chemical properties of the ligands, we found that the docking tools preferred compounds with different properties. FlexX calculated the lowest scores for ligands which possess 3 to 7 donors, 2 to 5 rings and a polar surface area of 50 to 210 2 while AutoDock prefers ligands with 1 to 3 donors, 5 to 7 rings and a polar surface area of 20 to 130 2. What both tools share is a preference for larger molecules with a high molecular mass which corresponds well to the reference ligands co-crystallized with the PDB structures. These findings clearly demonstrate that an application of several docking tools can enrich the candidate set and that parameters and target scenarios should be carefully chosen in advance if it is not possible to perform several experiments.

Identified compounds

Several strategies were employed to analyze the results including docking scores, ideal binding modes, and interactions to key residues of the protein. From the 500,000 ligands screened a few hundred compounds that can be tested in experimental laboratories should be identified. Three different classes of structures with thiourea, diphenylurea and guanidino scaffolds were identified to be promising hits. Diphenylurea analogues are already known to be micromolar inhibitors for plasmepsin (Walter Reed compounds, Jiang, S.; Prigge, S. T.; Wei, L.; Gao, Y. E.; Hudson, T. H.; Gerena, L.; Dame, J. B.; Kyle, D. E. New Class of Small Nonpeptidyl Compounds Blocks Plasmodium falciparum Development In Vitro by Inhibiting Plasmepsins. Antimicrob. Agents Chemother. 2001, 45, 2577-2584). This suggests that the overall approach is sensible for the discovery of inhibitors for plasmepsins. Guanidino analogues are likely to be a novel class of compounds, as they have not yet been reported as inhibitors for plasmepsins.


  • Poster Vinod Kasam
  • Presentations
  • Nicolas Jacq at WISDOM Open Day, Bonn, December 16th 2005 (Link)
  • Marc Zimmermann at WISDOM Open Day, Bonn, December 16th 2005 (Link)
  • Vinod Kasam at Dengue Docking Workshop, Basel, June 28th 2006
  • Antje Wolf at GridKa Summer School, Karlsruhe, September 14th 2006