Scientific Report



The workshop ‘Optimizing Drug Design’ was focused on the topic how novel multi-criteria optimization and decision making algorithms can be applied and adapted to the field of de novo drug discovery. This emerging field of research in cheminformatics holds high promises for the future as it will allow the chemical expert to perform a more focused search in the vast search space of possible chemical compounds. Moreover the integration of models, data, and semi-automatic search techniques may allow the expert to come up with new ideas for molecule designs.

The workshop brought together about 40 experts from more than ten different countries, working in academia or industry on optimization algorithms, cheminformatics and medicinal chemistry. It was one of the main intentions of the workshop to establish contacts and exchange between the experts from computer science and chemistry; on the one hand to improve the understanding of the drug design field by the computer scientist, and on the other hand for the medicinal chemists and cheminformaticians to learn about the state-of-the art approaches and recent trends in the field of optimization techniques. The response from the participants of the workshop indicated that this exchange was really fruitful and needed, and many new contacts were established. From this point of view the workshop was a big success.

Each day of the workshop was devoted to a general theme; starting with presentations on that topic and ending with a discussion session:

The purpose of the first day was to provide an overview of the field: A general discussion of the tools that are already available, their benefits and shortcomings and directions for future developments. The event started with a keynote presentation by C. Nicolaou, one of the pioneers in applying multicriteria evolutionary optimization in the domain of drug design. The subsequent presentation on the state of the art De Novo Design techniques in drug discovery by N. Brown set the scene for the subsequent interactive session: In a card-game the participants had to comment on their view of state-of-the-art, problems, goals, and promising approaches in optimizing drug design. The notes were collected on sheets of different colors and collected on a pinboard, so that each of the participants could form a picture on the current opinion in drug design. Talks by J. Holliday and P. Bonnet rounded up the first day with a discussion of novel techniques in virtual screening and conformational search.

The second day was focused on the topic of dealing with uncertainty in the models and in the goal formulations. The keynote presentation of C. Poloni and D. Di Stefano provided insights gained in studies on robust optimization achieved in collaboration between design optimization experts and the pharmaceutical industry. Later on, in the presentations on diversity based search and desirability indexes as well as the discussion of novel robust multicriteria optimization algorithms the state of the art of uncertainty handling in algorithms was discussed. Among others presentations were given by E. Zitzler and H. Trautmann, two leading experts in multicriteria algorithm design. In four subgroups the design of algorithmic approaches to deals with various types of uncertainties were discussed, and the results presented and discussed subsequently. One of the results was that often synergies between the different approaches can be exploited, i.e. by combining desirability functions with robust hypervolume-based optimization. Moreover the point of view of chemist on constraints such as the Lipinsky rule of 5 was discussed and further clarified.

The third day focused on the topic of ‘the role of the expert’. The interaction between the computer and the human expert is of major importance in the knowledge intensive domain of drug design, where a fully automated search is deemed to fail. After two presentations by A. Goldblum and O. Shir presenting new algorithmic ideas, the keynotes by I.C. Parmee and M. Seagull both emphasized on interactive search and they presented techniques on how to design human-centric design workflows for exploring chemical space. As on the previous day, an interactive session invited the participants to discuss this topic in subgroups and present the results in a summary session then. It was one of the results that optimization is often used as a learning tool, which requires additional features in the algorithmic design, such as explaining why a solution performs well or was rejected by an algorithms. Moreover, there were interesting ideas how to exploit the knowledge on possible chemical reactions in order to design more realistic transformations in search.

Besides the topic oriented discussion, the software demo and poster session was a highlight of the third day. The day was rounded up by a boat-cruise with dinner on the Kaag lakes north of Leiden, along meadows, boathouses, and windmills.

On the final day of the workshop, case studies for drug discovery, in which experts used existing tools to search for drugs on specific targets were in the center of interest. The talks held by PhD students in the field gave interesting insights on how the existing tools perform in practice and how they can be further improved. The summary discussion took a second look at some of the research problems stated on the first day in the light of the results of the results of the workshop. It was a consensus among most of the participants that by discussing these issues in a interdisciplinary and very focused forum brought the field one step further to the solution of these problems and made the participants more aware of what the real important questions are. For the future also the many new established contacts will help to achieve future progress.  

Presentations and abstracts of the participants, as well as the summary slides of the discussions are now available online for those who want to recapitulate the results. The scientific organizers want to express our gratitude to Corrie Kuster and Martje Kruk from the Lorentz Center, for their assistance in making the workshop a great event. Also we express our gratitude to the people working in the Pharma-IT platform for their assistance and contributions. Last but not least we acknowledge the financial support of TI Pharma and the Lorentz Center.


Michael Emmerich (LIACS, Leiden University, Netherlands)

Andreas Bender (LACDR, Leiden University, Netherlands)