OEChem Enabled: Data Mining Tools and Chemoinformatics Virtual Classroom

1
OEChem Enabled Data Mining Tools and
Chemoinformatics Virtual Classroom

• Norah MacCuish
• John MacCuish
• TJ ODonnell
• Tudor Oprea
• Mitch Chapman

Mesa Analytics Computing, LLC TJ ODonnell
Consulting Sunset Molecular, LLC Desert Moon
Consulting, LLC
2
Data Mining Tools

• Fingerprint Module programs reads SMILES strings
  and generates the 320 MDLI MACCS keys (or 164 of
  the earlier 166 public keys), Reoptimization of
  MDL keys for use in Drug Discovery, JCICS,
  42(6), 2002, 1273-1280. The programs are OEChem
  1.2 enabled.
• Grouping Module contains clustering algorithms,
  symmetric, asymmetric, non-overlapping and
  overlapping versions of Taylors algorithm. It
  also contains RNN hierarchical algorithms such as
  Wards, Group Average, Complete Link. The
  clustering algorithms contain their respective
  ambiguity indices. Measures program for
  similarity searching or for the generation of
  symmetric or asymmetric similarity matrices
  (sparse or full). It comes with various measures
  such as Tanimoto, Euclidean, Cosine, etc.
• Ambiguity Viewer is a Python TK application for
  viewing hierarchical results from the Grouping
  Module. OEDepict enabled.
• ChemTattoo program will produce a series of modal
  statistics and visualization useful in finding
  structural commonalities.  The data sets can be
  large or small (e.g., a specific group from a
  clustering).  The modal fingerprint of a data set
  is, loosely, the intersection of bits turned on
  for a given set of fingerprints.  This
  intersection can be parameterized to create a
  modal such that if a bit is turned on in the
  modal if for some percentage of the fingerprints
  (e.g., 90), that bit is turned on in the modal
  fingerprint. Via the MDL fingerprints above and
  atom counting, this program is OEChem 1.2
  enabled.

3
Chemoinformatics Virtual Classroom

•  
• At present there are no specific software tools
  for chemical information training in the U.S.,
  though there are several nascent chemometrics and
  chemical information university departments and
  curricula. A number of commercial software
  products used in the pharmaceutical and
  biotechnology industry are either too expensive
  or of limited utility for training in either
  academic or business settings. This poster will
  discuss converting Mesa Analytics Computing,
  LLC with OEChem inside into training tools for
  concept learning. These tools will serve the
  dual purpose in providing cheminformatics
  training tools for both academia and industry.
  By employing distance learning through a web
  delivery system, the training software will
  provide an effective, low cost solution for
  academic institutions, whether they are offering
  a single course to rural students in a remote
  setting, or an entire program in cheminformatics
  and a major urban university. In addition, such
  training tools will be very useful in industry
  settings with local area networks, where in a
  multidiscipline setting individuals need to
  receive training on the concepts employed by
  industrial chemoinformatics software an
  integral part of the drug discovery process for
  pharmaceutical and biotech industries.
•  
• Concerning the Chemoinformatics Virtual
  Classroom, our research results are based upon
  work supported by the National Science Foundation
  Small Business Innovative Research (SBIR) Program
  under Grant No. 0339360. Any opinions, findings,
  and conclusions or recommendations expressed in
  this material are those of the author(s) (Mesa
  Analytics Computing, LLC) and do not
  necessarily reflect the view of the National
  Science Foundation (http//nsf.gov).
• Abstract at https//www.fastlane.nsf.gov/servlet/s
  howaward?award0339360
•  

4
Ambiguity Viewer
5
Ambiguity Viewer
6
ChemTattooTM
The modal fingerprint of a dataset is, loosely,
the intersection of bits turned on for a given
set of fingerprints.  E.g., Fingerprints 010001
010100 010100 --------- 010000 Modal
fingerprint at 100 010100 Modal fingerprint at  
66 A simple example might be as follows 1.
Select a group of compounds, say a cluster from a
clustering of a collection of active
compounds.2. Generate the fingerprints for the
group, if they are not already generated.3.
 Input the fingerprints to ChemTattoo, specifying
the type of fingerprint (MACCS 166 or MACCS
320)4.  The output will contain the various
modal statistics and fingerprints mentioned
above, or the SMARTS patterns.5.  Input the
SMARTS patterns and original compound
fingerprints into a depictor with SMARTS matching
and atom/bond coloring. For an introduction to
modal fingerprints see, Stigmata  An Algorithm
to determine Structural Commonalities in Diverse
Data Sets , N. E. Shemetulskis, et al., JCICS,
1996, 36(4), 862-871.
7
HIVRT Non-nucleoside Inhibitors
RED Common modal features Black Not in
modal Threshold set to 1.0
8
Top Selling 200 Drugs
RED,GREEN, BLUE Common modal
features Black Not in modal Threshold set to
0.8
9
Acknowledgements

• OpenEye Scientific Software
• OEChem

• Sunset Molecular Discovery, LLC
• Wombat Database

• ChemAxon
• Marvin Tools