Molecular Informatics Department, BioFocus plc

On the 13^th March, the RSC’s Molecular Modelling Group held its second annual ‘Cutting Edge Approaches to Drug Design’ meeting.  Held in collaboration again with the Biological and Medicinal Chemistry Sector, the meeting attracted 127 delegates.  This was an increase on last year’s meeting, proving the event to be a great success for both groups.

The focus for this year’s meeting was to explore the leading technologies currently employed in key areas of the drug-discovery process, including biological target finding, virtual chemistry, lead optimisation, Absorption Distribution Metabolism and Excretion (ADME).  With the inclusion of some worked examples, the goal was to examine the current state of the art, highlighting key issues in the field.

Prof. Janet Thornton opened the conference with the keynote lecture, presenting her vision of the future of bioinformatics and its impact on drug discovery.  In it, she stressed the issue of managing the continually increasing flow of information from a wide range of sources, including sequencing, structure, transcriptome and proteome projects.  While information generated from these sources is critical to improving the drug discovery process, the real challenge will be in data integration.

David Parry Smith followed on this theme, but made the distinction between data, information and knowledge.  In particular, he emphasised that ‘more data’ does not mean ‘more knowledge’, highlighting the difference between manually curated database sources versus automated ones.  His implication was that a further challenge for the future would be in the efficient summarising of data.

Exploring lead discovery, Mike Hann examined the notion of “ligand complexity” versus the probability of finding a “hit” in HTS.  Using a very simplified receptor-site / ligand model, he demonstrated that the optimum ligand complexity is potentially much smaller than is observed in current libraries.  To finish, he argued that such smaller hit molecules would leave more room for the medicinal chemistry approaches in the hit to lead development.

 Alexander Alex followed the HTS theme, examining the issues with the current virtual screening technologies.  While they can identify hits, discrimination of inactive compounds is still very poor.  He suggested that a key aspect has been overlooked, i.e., the inclusion of negative interactions in scoring functions.  Training has traditionally been based exclusively on experimental data of high affinity compounds, where such interactions are characteristically under represented.

In the next talk, Han van de Waterbeemd drew attention to providing pharmacokinetic and drug metabolism data quickly in the HTS environment.  With the move to high throughput methods, data is increasingly being used in the development of predictive in-silico tools (cited as e-ADME).  While promising as a filter for virtual libraries, these approaches are still far from being fully validated.  His outlook was that significantly more data will need to be provided in the future to address the robustness of e-ADME tools

 Looking on to lead optimisation, Andrew Baxter discussed moving from screening ‘hits’ to lead compounds.  He reasoned that many lead optimisation studies fail to produce candidate drugs as a result of poor pharmacokinetic profiles, but that these characteristics could have been identified earlier in the ‘hit’.  To support this, he demonstrated key changes applied in the ‘hit-to-lead’ approach at AstraZeneca.

In the final session of the day, Jonathan Mason and David Selwood provided some worked examples, to illustrate some of the experiences and lessons learned in the drug discovery process.

Overall, the meeting successfully highlighted many of the key issues created by the new technologies in the drug-discovery process.