• Better, faster, smarter, …?
  • but does more yield better insight?
• How to manage data volume
  • keep track of it all
  • decide what is important
• How (or why) to use more complicated models?
  • need to develop new analysis tools
  • want to find new ways to visualize
• Be involved in process of data inquiry
  • share with colleagues (in real time)
  • reach beyond domain boundaries

• types of tools
  • algorithms: software tools
  • machines: hardware tools
• Your laptop
  • powerful for mid- to large-size projects
  • powerful communication tool
• Scaling up
  • Massive data: storage, management, access, sharing
  • Workflow steps: engine/CPU, organizing, scheduling

http://phytomorph.wisc.edu (Edgar Spalding Lab)

• Tools: Box, Google, GitHub
  • data, methods, ideas, results
• Research community
  • Arabidopsis
  • Switchgrass
  • Maize
• Cross-species systems
  • DNA sequence (genomics)
  • Phe-gen relationship (QTL, systems genetics)

• Your time
  • familiarity with tools
  • familiarity with data
• Comparing results
  • model selection on one dataset
  • several methods on one data set
  • multiple data sites

• genotype diagnostics
• Distribution at locus
• Scatterplots with symbols (QTL, env)
• Genome scans
  • LOD profile
  • Allele scans
  • SNP scans (GWA Manhattan plots)
• multiple traits
  • over time or space (Moore)
  • networks (small, not hair balls)
  • box- or dot-plots over conditions

Advances in measurement, design and analysis would be academic without advances in computational technology.

• faster machines -> faster throughput of more stuff
• methods translated into algorithms
  • open source: freely distrubuted, easy to study
• standalone programs
• packages in language systems (R or Python or Matlab)
• interconnectivity of algorithms and data resources

dangers of email-based collaboration

• trading large files back and forth (slow, not secure)
• nearly impossible to keep track of versions
• minor updates require repeat sending

modern approach: use email to notify collaborators only

• GitHub to share code & ideas with version control
• Box/DropBox & Google Drive to share documents
• iPlant to improve data access & processing efficiency

modular philosophy of layers to separate

• back-end: data and compute processing layer
• middle-ware: analysis methodology layer
• front-end: human interaction and data visualization layer

will enable overlapping communities to

• customize local use
• share data, methods & results with other communities
• off-load data handling & compute headaches

Jansen & Swertz (2007)