ChemBank Building a Public Web Resource Using Daycart

1
ChemBankBuilding a Public Web ResourceUsing
Daycart

• Erik Brauner
• Head of Chemical and Biological Computing
• Harvard Institute of Chemistry and Cell Biology
• Eli and Edith L. Broad Institute
• June 10, 2004

2
The Institute of Chemistry and Cell Biology (ICCB)

• The ICCB is an academic small molecule screening
  facility located at Harvard Medical School with
  the goals of
• Enabling academic labs to perform high throughput
  chemical screens
• Creating small molecule libraries for screening
• Advancing the field of Chemical Genetics
• Creating a public database ChemBank

3
http//ChemBank.med.harvard.edu

• The Dual Roles of ChemBank
• To handle internal needs
• Support tools for high throughput screeners
• Support for chemists and library synthesis
• As a public web resource
• Freely available assay data and information on
  compounds relevant to chemical genetics
• Had to satisfy the needs of both chemists and
  biologists.

4
(No Transcript)
5
ChemBank Status

• Publicly available at chembank.med.harvard.edu
• gt900,000 structures in the database
• gt5,000 known bioactives with annotation in the
  database
• Selected assay data also available
• Supports similarity queries, and substructure
  searching.

6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13

• Structure Sources
• Commercial libraries
• Outside databases
• Curated structures
• DOS Libraries
• Virtual library files with
• undecoded structures (no
• plate mapping), decoding
• tag pattern, building blocks,
• reagents

14
(No Transcript)
15
SD File Handling
SD file
mol2tdt.sh (contrib code)
TDT file
custom PERL script
This becomes the official load record
XML file
16
A Simple Structure Table

• create table COMPOUND (
• id number not null,
• smiles varchar2(4000),
• molecular_weight float,
• molecular formula varchar2(20),
• primary key (id)
• )
• create index COMPOUND_idx1
• on COMPOUND(smiles)
• indextype is cdcischem.ddexact
• create index COMPOUND_idx2
• on COMPOUND(smiles)
• indextype is cdcischem.ddblob

needed for exact smiles matching
needed for fingerprints to support similarity
17
Basic Manipulations

• Inserting Compounds
• insert into COMPOUND
• (id, smiles, molecular_weight, molecular_formula)
• values ( 1, smi2cansmi(CCC, 1), smi2amw(CCC),
  smi2mf(CCC))
• Similarity Search
• select id, tanimoto(smiles, OCO) as similarity
• from COMPOUND
• where tanimoto(smiles, OCO) gt 0.8
• order by similarity desc
• Substructure Search
• select id
• from COMPOUND
• where contains(smiles, OCO) 1

18
(No Transcript)
19
Structure Loading
Are These Structures of the Same Compound?
?
20
(No Transcript)
21
Salt Stripping In Daycart

• Daycart 4.8.2 supports salt stripping via the
  function vcs_desalt(smiles, iso, class) which
  works in conjunction with a built in salt table.
• ex
• insert into salt values (Sodium, Na, 0 ,
  NULL)
• VCS_DESALT(Na.c1ccccc1, 0, 0)

22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
Normalization in Daycart

• Nitro and azide normalization can be achieved
  easily using reaction smirks.
• Ex
• 1N2(O3)O4gtgt1N2(O3)O-4

26
Normalization In Daycart

• Daycart 4.8.2 supports normalization via the
  function vcs_normalize(smiles, iso, class) which
  works in conjunction with a built in transform
  table.
• ex
• insert into transform values (Nitros,
  1N2(O3)O4gtgt1N2(O3)O-4
  , FORWARD, 0 , NULL)
• VCS_NORMALIZE(CCN(O)(O), 0, 0)

27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
Acknowledgements

• ICCB Informatics Group
• Jeremy Muhlich
• Jason McIntosh
• Carol Chang
• Andrew Lach
• Justin Klekota

• ICCB Chemistry Group
• John Tallarico
• Jared Shaw
• ICCB Screening Group
• Caroline Shamu
• Nicky Tolliday

National Cancer Institute Tudor Oprea
Daylight