HOME- R&D
- Bryn Mawr Conference
- Program & Schedule
- Location
- Screening Workshop
- Call for Contributions
- Bursary Awards
- Poster Session
- Sponsors & Exhibitors
- Program 2007
- ADMET
- Fragment-based Design
- SBDD
- Screening
- Austin, C
- Delfaud, F
- Hawkins, P
- Hert, J
- Langenaeker, W
- Liebeschuetz, J
- Reid, D
- Ruvinsky, A
- Stahl, G
- Taylor, R
- Zsoldos, Z
- Structural Biology
- ADME/Tox Forum
- Program 2006
- Photo Gallery
- Hyderabad Conference
- Workshops & Training
- Program
- Exhibition
- Registration
- Jobs
- Contact
- Schedule
|
|
|
|
|
|
|
Jérôme Hert obtained his BSc in chemistry and MSc in chemoinformatics from the Louis Pasteur University in Strasbourg (France). He spent his MSc internship at the Novartis Institutes for BioMedical Research in Basel (Switzerland) working on high-throughput screening data analysis after which Novartis offered to sponsor his PhD. He hence joined the chemoinformatics group of Professor Peter Willett at the University of Sheffield (UK) in December 2002 and developed new algorithms for similarity searching using one or several query molecules; he obtained his PhD in January 2006. He is now undertaking postdoctoral training at the University of California, San Francisco in the group of Professor Brian K. Shoichet where his research aims at relating proteins through the statistical similarity of their ligands. He was awarded a European Marie Curie Outgoing International Fellowship in May 2007.
|
|
Pharmacological networks of proteins derived from the similarity between their ligand-sets.
Jérôme Hert, University of California, San Francisco
Increasing attention is given to approaches that relate proteins through the similarity of their ligands. The idea consists of using the chemical similarity among ligand-sets as a proxy to the pharmacological similarities between protein targets. Chemical mapping of pharmacological relationships relates targets in a formal network where the edges represent their similarity to one another. These networks complement the more familiar bioinformatics networks and reveal relationships between targets that would be obscured based on sequence or structural similarity alone. Here we quantify the differences between chemoinformatics and bioinformatics networks and explore the consistency of these differences. We also investigate how dependant the chemoinformatics networks are on the similarity method and the molecular representation. Specifically, we consider two different approaches to compare the ligand-sets (a statistical Blast-like method based on the similarity of pairs of ligands and a model-based method based on Bayesian inference) and seven different representation of chemical information for the chemoinformatics networks versus the PSI-Blast similarity between sequences for bioinformatics networks. We find there is no significant similarity between a network of drug targets based on sequence identity and one based on ligand-set similarity, irrespective of how we compare or represent the ligands. Conversely, the ligand-set relationships are largely preserved for multiple ways of representing chemical information, and are robust even to the particular choice of ligands within these sets. Most importantly, the pharmacological relationships inferred from ligand-set similarity are relevant and lead to testable predictions. Chemoinformatics networks of receptors may be used to predict off-target effect, polypharmacology, side-effects and drug-repurposing.
|
|
|
|
|
|
|
|
|
