=== James Taylor, Galaxy platform for accessible and reproducible research ===

- Evolution of biological functional elements and speciation events
- In DNA, substitutions, insertions, deletions
- Evolutionary constraint prevent changes in functional regions

- Lots of development in genomic data warehouses and visualization methods
- Many computational methods, too; lots of methods/application papers published with working implementations
- Typical workflow: developer provides scripts, grad student runs test cases from Excel worksheet or perl script

- Galaxy: an open source framework to integrate computational tools and databases into a cohesive workspace
- Web interface for all analysis operations
	- Register tools, access others by proxy
	- Integrate existing command-line tools by describing them in XML
	- Use template to generate command line from parameter values
- Include tests, support for multiple "tracks"
- Histories: complete records of complex analysis; can be easily shared, rerun
- Create workflows
- Python 2.4, WSGI web framework, SQLAlchemy, jQuery
- Can be run locally, or scale with mod_proxy and pbs-python job dispatching

=== Gregory Benson, River: reliable virtual resources ===

- A Python framework for parallel and distributed programming
- River core: discovery, process naming and creation, message passing, state management
- River extensions: RPC/RMI, Trickle, MPI, MapReduce

- Parallel programming still hard, will have to use multiple cores, larger clusters
- Find the next set of programming models
- Current practice for development cycle
	- Specify -> Prototype -> Get feedback
	- Takes months, and early decisions are hard to undo
- River purpose
	- Extend Python's rapid development capabilities to parallel systems
	- Facilitate short design/implementation cycles

- River concepts:
	- Virtual machines (Python + River Core)
	- Virtual resources
		- Named
		- Code, data, thread, message queue
	- Coordinated checkpointing
	- Super flexible messaging

- https://www.cs.usfca.edu/river/

=== Eric Jones, Block and Contexts: exploring scientific problems and algorithms ===

- Scientists know how to model problems in software, but exploring them is harder
- Code written by scientists usually has few objects, hardcoded parameters, prints out a graph
- Make interactive by adding traits to parameters, array output... not that simple
- Lots of overhead for graphical applications

- Model analysis:
	- What if (fixed parameter value)
	- Parametric studies (parameter ranges)
	- Monte Carlo
	- *Inverse problem*

- Blocks and context
	- Execute a block (code) in a context (parameter values)
	- Can do it in python with "exec code in {}, context"
	                                string      dictionary
	- Shadow contexts in what-if experimentation to reuse data
	- Implement in visual interface