exploRase

1
exploRase

• Exploratory Data Analysis Software for Systems
  Biology

Michael Lawrence Co-Major Professors Di Cook and
Eve Wurtele
2
Outline

• Background
• Exploratory Data Analysis
• Software Foundation
• Systems Biology
• Introduction to exploRase
• Explore some data (demo)
• The Future
• Acknowledgments

explo R ase
3
Exploratory Data Analysis

• Exploratory Data Analysis (EDA) is the art of
  being a data detective.
• Instead of working to confirm a single
  hypothesis, we approach the data with a loose set
  of questions and/or suspicions and remain open to
  the unexpected.
• We look for trends and patterns, as well as
  deviations from those expected.
• Interactive graphics facilitate fluid data
  exploration and link different views of the data
  by visual cues.

4
Software Foundation

• R Platform for statistical computing. Provides
  flexible data analysis via scripts.
• GGobi Leading open-source interactive graphics
  software. Mature and actively developed. Based on
  the GTK toolkit.
• Rggobi2 Link between R and GGobi. R's scripted
  analysis meets GGobi's interactive graphics.
  Overhaul of previous version.
• RGtk2 Binding between R and GTK (open-source
  GUI toolkit). Allows GUI's to be written in R and
  integrated with GGobi's GUI via Rggobi2.

5
Systems Biology

• The highly computational study of biology as a
  complex (biochemical) system.
• Based on the analysis and modeling of metabolic,
  regulatory, and protein interaction networks.
• Networks are often inferred from or analyzed in
  conjunction with high-throughput data
  (transcriptomics, proteomics, metabolomics)
  describing system states.

6
exploRase is...

• An RGtk2-based graphical user interface that
  makes advanced statistical tools like R and GGobi
  more accessible to a systems biologist.
• Facilitates loading of experimental data into R
  and GGobi.
• Will serve as a front-end to BioConductor (R
  software for Bioinformatics).

7
Let's go exploring...

• Dataset Arabidopsis transcriptomic data from 8k
  Affymetrix chips.
• 2 Genotypes WT and bio1 (deficient in biotin
  synth)
• Half of samples given biotin, total of 4
  treatments.
• 2 replicates
• Data Source Basil Nikolau's lab

8
The Future

• Integrate with an interactive network
  visualization tool (for example Cytoscape)
• Add more analysis methods and data aggregation
  utilities, with emphasis on interactivity
• Refine user interface
• Context sensitive help and wizards
• Integration with other MetNet tools

9
Acknowledgments

• GGobi Debby Swayne (ATT), Di Cook, Duncan
  Temple Lang (UC-Davis), Andreas Buja (Wharton),
  Heike Hofmann, Michael Lawrence, Hadley Wickham
• Rggobi2 Hadley Wickham (R), Michael Lawrence
  (C), Original Duncan Temple Lang, Debby Swayne
• RGtk2 Michael Lawrence, Original Duncan Temple

Money MGET Fellowship, NSF Arabidopsis 2010