On the Path to Drug Discovery: New Small Molecule Resources

1
On the Path to Drug DiscoveryNew Small
Molecule Resources

• Michel Dumontier, Ph.D.
• Carleton University

2
The Path to Drug Discovery

• There are essentially six steps in drug
  discovery
• disease selection
• target identification
• lead compound identification
• lead compound optimization
• pre-clinical trial
• clinical trial

3
In essence, we need to

• Identify proteins that bind small molecules.
  These are potential targets and their small
  molecule hits.
• Reduce the list of targets by identifying
  essential genes.
• Reduce the hits by considering specificity of
  interaction.

4
Protein Small-Molecule Interaction Database

• 23,000 non-redundant protein small-molecule
  interactions.
• Derived from PDB structure database
• Filtered for crystallographic symmetry, buffer
  agents, non-biologically interesting small
  molecules
• Captured in the 3DSM division of the Biomolecular
  Interaction Network Database (BIND)
  http//bind.ca

Biopolymers. 2001-2002 61(2)111-20
5
Small Molecule Curation
Internal Links BIND Records Binding Domains
External Links PDBSum HIC-up MMDB PDB KEGG PubC
hem EcoCyc
Taxol
Approved for treatment of breast cancer (1994)
6
SMID Mapping small molecule interactions to
domains
Taxol binding conservation in Tubulin/FtsZ domain
7
Conserved Domain Ligand Visualization
Small molecule and contacting ligands are
automatically rendered/colored Domain
Purple Ligands Green Taxol Spacefill
8
Use Cases for SMID

• Domain Studies
• Work with small molecule binding domains
• Conservation of binding site across domain
  members
• Structural Genomics
• Domain/ligand/binding site identification
• Some ligands go over domain boundaries
• Domain family entry into PDB structures
• Quickly identify candidate co-crystalization
  ligands

9
SMID-BLAST

• Enables users to predict small-molecule binding
  sites in proteins for which a crystal-structure
  has not yet been determined.
• Maps binding sites to the query protein from
  experimentally determined interactions.
• requires a hit to a conserved domain.
• Freely available
• Web interface
• http//smid.blueprint.org
• Standalone tool
• ftp//ftp.blueprint.org/pub/SMID/tool/

10
Protein Sequence
Alignment to conserved small-molecule binding
domains
Evaluate binding site conservation
domain small- molecule interactions
cluster binding sites from all domains
List of small-molecule binding sites with
confidence scores
11
Validation Summary

• 1652 experimentally determined protein
  small-molecule interactions from PDB
• Predictions
• Ligand correctly predicted in 62 of cases.
• 25 of the ligands also obtained the best ligand
  score.
• 73 of interactions were predicted with gt80
  binding site coverage
• This is very good, as the test set is not
  comprehensive
• we dont have a set of all possible ligands to
  each protein crystal structure.
• we can only use exact small molecule matches (not
  similar molecules, e.g. ATP vs ATP-gamma-S)

12
SMID-BLAST ExampleHIV Integrase

• Mediates integration of a copy of the viral
  genome into the host DNA.
• Lack of mammalian counterpart makes for good drug
  target.
• There is no solved structure as of yet, but has
  known Zn binding domain, a catalytic core and
  DNA-binding domain.
• Use SMID-BLAST to make short list of small
  molecules that interact with the integrase
• could lead to pharmacological studies to
  determine inhibition.

13
Small molecules predicted to bind to HIV Integrase

• 4 clustered binding sites
• 4 small molecules
• 5 ions
• 100 5CITEP
• known inhibitory ligand
• TTA - tetraphenyl-arsonium
• known interactor, but doesnt inhibit
• Y3
• known interactor with Avian Sarcoma Virus
  integrase
• New compounds identified from pharmacophore search

14
Integrase Binding Site Mapping

• Each hyperlinked residue indicates binding site
• Clicking on link shows coverage of
  domain-specific binding sites, and a link to the
  SMID, PDB, BIND record that supports the
  interaction.

15
A genomic context

• SMID Genomes bridges the gap between structural
  proteomics and genomics
• Small-molecule interaction predictions for
  proteins of all genomes
• Allows for comparative analysis of small-molecule
  binding profiles

16
Application area of Prokaryotic Genome
Projects
Source http//www.genomesonline.org
17
SMID-Genomes

• Ran SMID-BLAST for all protein sequences
• 10.5M small molecule binding interactions
• Conservative threshold
• Added genomic context
• 9.4M interactions across 385,000 proteins from
  1558 completely sequenced genomes
• 50 protein coverage for archae,bacteria
  eukaryotes
• 35 for viruses/phages
• Search/Browse/Compare predicted small molecule
  interactions in a taxonomy specific manner

18
Search
Browse
Compare
19
Browsing SMID-Genomes
20
Protein view lists all proteins with putative
small molecule interactions
Click to list small molecules predicted to bind
to this protein
Click to view small-molecule binding sites on
genomic proteins
Limit the search results by a term!
21
Small molecule view includes picture and link to
small molecule summary page
Click to view domain-specific interactions
Click G to list all genomes having a
interaction with this small molecule
22
Domain view lists the conserved domains used to
make the small molecule predictions
Click G to list all genomes having a
interaction with this domain
23
Multi-Genome Small-Molecule Binding Profile
Comparison

• Malaria
• a disease that directly impacts 300-500 million
  people worldwide and is a prominent economic and
  social problem in the developing world
• Browse small molecules that target proteins in
  humans, Plasmodium falciparum, the parasite that
  causes malaria and Anopheles gambiae, the
  mosquito that carries the malarial parasite.

24

• Fosmidomycin is an antibiotic isolated from
  Streptomyces lavendulae operates as a potent
  inhibitor of DOXP reductoisomerase, a key
  enzyme of the alternative pathway of isoprenoid
  synthesis.
• It is known that the malaria parasite Plasmodium
  is dependent on this pathway, because it lacks
  the primary isoprenoid synthesis pathway.

25
Small-Molecule Specificity

• Hits 74 of bacteria
• For eukaryotes, hit to rat, but only plant has
  ortholog of reductoisomerase

26
Text Query SARS
27
Essential Genes

• Putatively identify essential genes with small
  molecule interactions and find even better
  targets.
• Database of essential genes http//tubic.tju.edu.c
  n/deg/
• Haemophilus influenzae. 2002. Proc Natl Acad Sci
  U S A 99 966-971.
• Mycoplasma genitalium. 1999. Science 286
  2165-2169.
• Staphylococcus aureus. 2001. Science 293
  2266-2269.
• Vibrio cholerae. 2000. Nat Biotechnol 18
  740-745.
• Bacillus subtilis. 2003. Proc Natl Acad Sci U S A
  100 4678-4683.
• Staphylococcus aureus. 2002. Mol Microbiol.
  431387-400.
• Streptococcus pneumoniae. 2002. Nucleic Acids
  Res. 303152-62.
• Helicobacter pylori. 2004. J Bacteriol.
  1867926-7935.
• E. coli http//www.shigen.nig.ac.jp/ecoli/pec/inde
  x.jsp
• Yeast - MIPS http//www.mips.biochem.mpg.de/proj/y
  east
• Map essential genes to other genomes by ortholog
  assignment
• Identify best reciprocal BLAST hits between
  organisms
• Inparanoid algorithm
• Sonnhammer et al. 2001. JMB, 3141041-1052.

28
Identify small molecules that exclusively
interact with mycoplasma urethral pathogens
Chicken Urethral (Men) Swine Non-pathogen Cattle H
uman (resp.)
29
FM2 nucleoside inhibitor acts on an essential gene
30
Converting Small Molecules hits to Lead Compounds

• So far, we have been using small molecules in
  PDB
• Expand small molecule library with NCBIs PubChem
  650,000 compounds
• Display small molecule hits with
  chemical/biological significance.
• Chemical Ontology

31
Chemical Ontology

• 233 Chemical Functional groups
• Simple organization illustrations
• Functional groups automatically assigned by
  checkmol computer program
• Can identify molecules having two or more
  functional groups
• Semantic similarity algorithm to compare small
  molecules on shared terms
• Applied to internal SMDB PubChem

32
Chemical Ontology applied to PubChemamine
functional groups
parent
child
siblings
33
Union of functional groups
34
(No Transcript)
35
Next Steps

• Compare PDB small molecules to PubChem compounds
  with semantic similarity algorithm
• Filter similar molecules by drug-like properties
• Lipinski rule of 5 molwt lt 500, logp lt 5.0, lt5
  h-bond donors, lt10 h-bond acceptors
• Add any available biological assay information
  (deposited in Pubchem)

36
Future Directions

• Identify diseases that may be affected by binding
  site mutations in genes described in the OMIM
  database.
• Do known variations disrupt binding?
• Identify small molecule binding sites that are
  affected by non-synonymous SNPs in
  disease-related proteins.
• Experimentally investigate via wet-lab
  collaborations

37
Conclusions

• New and freely available resources for drug
  discovery
• A software program that annotates small-molecule
  binding sites to protein sequences with no known
  structure.
• A platform for drug discovery that considers
  target exclusivity and compound specificity
  across all genomes.

38
Acknowledgements

• Christopher Hogue Blueprint PI
• SMID/SMID-BLAST
• Howard Feldman
• Kevin Snyder
• Small Molecule Group
• Susan Ling
• Brian Parker
• Roberta Stasiuk

• BIND
• Marc Dumontier
• Anthony Hrjovic
• Shawn Konopinksy
• John Salama
• IT
• Sam Sgro
• Funding Agencies
• Genome Canada
• Genome Ontario
• CIHR
• ORDCF