CICC Chemical Compound Mining Workflows

1
CICC Chemical Compound Mining Workflows

• Jungkee (Jake) Kim
• Community Grids Laboratory

2
A Workflow for Big Red Demo I
PubMed Abstracts
OSCAR3
SMILES Extraction
Converting the format
XML files
Text files
SMILES
Molecular Quantum Mechanics
Converting to pictures
Generating HTML script
SDF files
SDF files
POV, JPG files

• Big Red is one of fastest supercomputers
• Mining chemical compounds found on research paper
  texts and showing them in 3D graphics

3
A Workflow for Big Red Demo II
Final HTML pages
4
A Workflow for Big Red Demo III

• PubMed abstracts
• 555,007 PubMed abstracts of 2005 2006 (part)
  R. Guha
• 1,000 abstracts per node distributed (Simple
  parallelism)
• 511 nodes X 1,000 input abstracts used for the
  demo
• OSCAR3
• A Cambridge tool which extracts chemical
  information from text and produces an XML
  instance highlighting the chemical information
• Used a revised version for convenient batch
  processing (some incompatibility to BigRed
  architecture)
• SMILES extraction
• Extracting SMILES elements from OSCARs XML
  output files
• Unique SMILES list within a batch

5
A Workflow for Big Red Demo IV

• Generating 3D formats K.
  Gilbert
• Converting from SMILES to SDF format
• Molecular Mechanics program mengine (MM
  engine)
• No Quantum Mechanics (QM) in the demo
• Converting 3D formats to pictures J. N.
  Huffman
• Persistence of Vision Raytracer (POV-Ray)
  converting SDF to POV
• Another program which converts the POV files to
  JPEG format
• Generating HTML script
• Showing those graphic files in an HTML page

6
Bigger Picture for the Workflow
NIH PubMed Database
OSCAR Text Analysis
Toxicity Filtering
Cluster Grouping
Docking
Initial 3D Structure Calculation
High Throughput Screening (HTS) Data Organization
and Flagging
Molecular Mechanics Calculations
NIH PubChem Database
Quantum Mechanics Calculations
Big Red Demo
IUs Varuna Database
POV-Ray Parallel Rendering