Predicting Biomolecule Interactions

1
Predicting Biomolecule Interactions

• Joel R. Bock and David A. Gough
• Department of Bioengineering
• and
• The von Liebig Center for Entrepreneurism and
• Technology Advancement
• University of California San Diego
• Spring, 2005

2
The Problem

• Molecular interactions are the basis of biology
• The number of interactions is enormous
• Need to focus on the most important ones

3
Our Solution

• Pattern recognition software for prediction of
  molecular interactions
• Works with DNA, RNA, proteins, small molecules
• Can identify new ligands or
• new receptors for existing ligands

4
How does it work?

• Pattern recognition algorithm is trained on
  known interactions
• Algorithm is then applied to novel chemistry
• 3-D protein structure is not needed
• uses only gene or AA sequence
• Java-based software runs on a desktop

5
Benefits of the Technology - 1
Software

• Focuses searches on new drugs for specific
  targets
• Reduces the number of wet experiments
• Can be used for design of high- throughput or
  secondary screening
• Interactions suggest protein function

6
Benefits of the Technology - 2
Identifies Novel Ligands

• Over 110 novel ligands predicted for 55 orphan G
  protein-coupled receptors (oGPCRs)
• Ligands from NCI public database
• Patents pending for use of ligands

7
Benefits of the Technology - 3
Identifies New Receptors

• Olanzapine an antipsychotic drug with known
  high-affinity binding to non-orphan GPCRs
• Predicted active to highly active for all
  oGPCRs
• Correctly predicted no activity in other cases

8
Benefits of the Technology - 4
Finding New Targets

• Screened high-affinity CNS nicotinic receptors
• 9 out of 17 correct hits within experimental
  range
• Demonstrates target-finding capability

9
Status of the Technology

• Six peer-reviewed publications
• Two pending patent applications
• Method validated in multiple ways
• Software implemented in user-friendly GUI

10
What Can We Do For Your Company?

• Demonstrate program on your real or test data
• User feedback to improve model further
• Beta test your known compounds under NDA for
  further validation
• Follow-up on leads for primary or secondary
  screening ? Cost Savings
• Collaboration and licensing

11
The technology is described in peer-reviewed
publications

• Predicting Protein-Protein Interactions from
  Primary Structure," Bioinformatics 17 (2001),
  1-6.
• Whole-Proteome Interaction Mining,"
  Bioinformatics 18 (2002), 1-10.
• "A New Method to Estimate Ligand-Receptor
  Energetics,
• Molecular and Cellular Proteomics 1 (2002),
  904-910.
• In silico Proteomics Predicting Interactions
  from Sequence, in Handbook of Proteomics, P.
  Michael Conn, ed., pp. 193-222 (2003) Humana
  Press, Totowa, NJ.
• In silico Biological Function Attribution A
  Different Perspective," Biosilico 2 (2004),
  30-37.
• Virtual Screen for Ligands of Orphan G-Protein
  Coupled Receptors, Journal of Chemical
  Information and Modeling, in press.

12
Questions? Comments?