33
Modeling of Biomolecular Systems Interactions, Dynamics, and Allostery Session VI Abstracts
Data Mining Large-Scale Bioactivity Datasets to Find Patterns in Ligand Recognition
John Overington
.
EMBL-EBI, Hinxton, United Kingdom.
We have built a large-scale open data resource, ChEMBL; this contains in excess of 1.4 million
compound structures, and over 10 million associated bioactivities. Where possible the data is
linked to molecular targets, and further annotation performed to provide deeper indexing and
organisation of the data. This has become an important resource for the community in
developing data-driven approaches to a number of important problems in drug design and safety,
including predicting targets and 'off-targets' for novel compounds, quantifying the drug-likeness
of compounds, and in the design of novel bioactive molecules. This data has been recently
complemented by a new resource, SureChEMBL, containing automatically text-mined data from
patent sources, signficantly increasing coverage of chemical and target space and diversity. The
presentation will present the principles in the organisation and features of ChEMBL, and then
some approaches addressing the difference between allosteric and non-allosteric ligands, target
predictions from phenotypic data, and finally analysis of polypharmacology - the binding of a
ligand to many components of a cell.
Combating Drug Resistance: Lessons from the Viral Proteases of HIV and HCV
Celia Schiffer
.
UMASS Medical School, Worcester, MA, USA.
Drug resistance negatively impacts the lives of millions of patients and costs our society billions
of dollars by limiting the longevity of many of our most potent drugs. Drug resistance can be
caused by a change in the balance of molecular recognition events that selectively weakens
inhibitor binding but maintains the biological function of the target. To reduce the likelihood of
drug resistance, a detailed understanding of the target’s function is necessary. Both structure at
atomic resolution and evolutionarily constraints on its variation is required. “Resilient” targets
are less susceptible to drug resistance due to their key location in a particular pathway. This
rationale was derived from our lab’s experience with substrate recognition and drug resistance in
HIV-1 protease and Hepatitis C (HCV) NS3/4A. Both HIV-1 protease and HCV NS3/4A
protease are potentially “resilient” targets where resistant mutations occur outside of the
substrate binding site. These principals are likely more generally applicable to other quickly
evolving diseases where drug resistance is quickly evolving.
