Transcript BioVLAB-Microarray: Microarray Data Analysis in Virtual Environment

BioVLAB-Microarray: Microarray Data
Analysis in Virtual Environment
Youngik Yang, Jong Youl Choi,
Kwangmin Choi, Marlon Pierce,
Dennis Gannon, and Sun Kim
School of Informatics
Indiana University
CONTENTS
•
•
•
•
•
•
•
•
Introduction
Approach
Related Works
Microarray technology
System Architecture
Experiments
Conclusion
Demo
INTRODUCTION
• Analysis of high throughput microarray experiment
• Performing microarray analysis is a demanding task
for biologists and small research labs
• Computing infrastructure issue
– Computationally intensive
– Nontrivial to integrate various bioinformatics applications
• Exploratory data analysis issue
– Multiple tasks in a single batch
– Repetitive execution
APPROACH
• On-demand computing resources
• A suite of microarray analysis applications
• Reconfigurable GUI workflow composer can alleviate
technical burden
– Well defined workflow can be repetitively used
• Web portal
• Reusable, reconfigurable, high-level workflow
execution workbench powered by computing clouds
for microarray gene expression analyses
RELATED WORKS
• Efficient and user-friendly workflow
composers and execution engine
– SIBIOS, BioWBI, KDE Bioscience
• Distributed and heterogeneous computing
resources + Workflow system
– Taverna, Triana, Kepler, GNARE, RENCI-Bioportal
MICROARRAY TECHNOLOGY
• A subset of genes is expressed
corresponding to environmental
changes and its changing needs
• Dynamics of cell activity
• Measure gene expression levels of
hundreds of thousands of genes
within a cell
• Usage
– Function prediction: Guilt by
association
– Interaction: Co-expression of genes
in transcription networks reveals how
they interact.
– Drug discovery: Identify genes
related to certain disease and detect
effectiveness of new drugs
Source: www.liv.ac.uk/lmf/about_microarrays.htm
RESEARCH GOALS
• Gene expression analysis
– Search for similar patterns of genes
• Similar patterns of gene may reveal the function of a gene with unknown function
– Extraction of differentially expressed genes
• Statistical evaluation
– Clustering
• Protein function prediction
• Genes with similar expression may need to be studied as a group
– Component analysis
• Hidden structure of expression patterns may be revealed
• Expression network analysis
– Expose hidden structures
– Protein-protein interaction (PPI) network analysis
• Central issue: key role in understanding how a cellular system works
• Modularity in structure in a network may reflect higher-level functional
organization of cellular components
MICROARRAY ANALYSIS COMMON
TASK
• Output of a task can plugged into another task
• Repeat the same set of tasks with small
changes of parameters
SYSTEM ARCHITECTURE
• Workflow composer and execution engine
• Application services
• Web portal
Workflow Composer
& Execution
Execute
Manage Data
Web Portal
Create
Application
Services
WORKFLOW COMPOSER &
EXECUTION ENGINE
• Introduced in the scientific communities to
execute a batch of multiple tasks
Input
• Enables repetitive tasks easily
• Directed acyclic graph
– Node: application to execute
• Starting node: input
• End node: output
Task A
Task B
Task C
– Edge: a flow of data
Output
XBaya
•
•
•
•
GUI Workflow composer and execution engine
Developed at IU
Drag-and-drop compose from workbench
Monitor status of workflow execution
Workbench
Panel
Application
Information
Panel
Workflow
Composer
Panel
Drag-and-drop
Monitor
Panel
APPLICATION SERVICES
• Interoperability among applications can be achieved
by Application Services
• Generic Service Toolkit (Gfac)
– Gfac converts command-line bioinformatics application
into a web service
• On-demand computing resources
– Amazon Elastic Computing Cloud (EC2)
• Remote storage services
– Amazon Simple Storage Services (S3)
– Microsoft Application-Based Storage
User
• Develop a command line app.
Admin
• Install the app. in Amazon EC2
• Let the app. store any output to
Amazon S3 / Microsoft
Application-Based Storage
• Make a virtual machine image
• Register the app. by using Gfac
User
BioVLAB APPLICATION
DEVELOPMENT PROCEDURE
• Instantiate EC2 and run the app.
by using XBaya
Gfac Registration form
(Gfac user manual)
WEB PORTAL
• Adiministrator
– Management of
registered applications
by Gfac registry portlet
– User management and
access control
• User
– access of stored data
• Built by Open Grid
Computing
Environments (OGCE)
ANALYSIS RESOURCES
•
•
•
•
•
•
R: statistical learning
Bioconductor: microarray analysis
Data acquisition: NCBI GEO Microarray DB
Similar expression pattern: correlation
Differentially expressed gene: limma package
Clustering: K-means, hierarchical clustering, QT clustering,
biclustering, Self organizing map (SOM)
• Component Analysis: principal component analysis (PCA) and
Independent component analysis (ICA)
• Network: Database of Interacting Proteins (DIP), Perl Graph
package and GraphViz
EXPERIMENT
• Data set: GDS38
– Remotely retrieved from the NCBI GEO database
– Time-series gene expression data to observe cell
cycle in Saccharomyces cerevisiae yeast genome.
– 7680 spots in each 16 samples
– Each sample was taken every 7 minutes as cell
went through cell cycle.
• Expression analysis
• PPI network analysis
EXPERIMENTS
CONCLUSION
• Microarray data analysis in virtual
environment
• Coupling computing clouds and GUI workflow
engine
• Effective system design for small research labs
FUTURE WORKS
• Integration of more packages and analyses
• A system of great flexibility
– Integrate various high throughput data
• Microarray, mass spectronomy, massively parallel
sequencing, etc
– Integrate various computing resources
• Clouds, grid, and multi-core PCs
– Integrate various public resources
• NCBI, KEGG, PDB, etc
SCREEN SHOTS
S3 BROWSER
EC2 ACTIVE INSTANCE
WORKFLOW FOR CLUSTERING
INPUT PARAMETERS
WORKFLOW EXECUTION
DATA ACQUISITION
SUBSET EXTRACTION
CLUSTERINGS
WORKFLOW TERMINATION
EXPERIMENT RESULT
DOWNLOAD FILE
HEATMAP FOR K-MEANS
CLUSTERING
ACKNOWLEDGEMENT
• The work is partially supported by NSF MCB
0731950 and a MetaCyt Microbial Systems
Biology grant from Lilly Foundations.
• Extreme Computing Group at IU
– Suresh Marru, Srinath Perera, and Chathura
Herath
Thank You