08.02.2013 Views

New Statistical Algorithms for the Analysis of Mass - FU Berlin, FB MI ...

New Statistical Algorithms for the Analysis of Mass - FU Berlin, FB MI ...

New Statistical Algorithms for the Analysis of Mass - FU Berlin, FB MI ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

152CHAPTER 6. PROTEO<strong>MI</strong>CS.NET - PRODUCT-ORIENTED CASE STUDIES<br />

Organization <strong>of</strong> data and results: Often a project lacks sufficient organization,<br />

documentation and storage <strong>of</strong> source and resulting data or parameters<br />

used in different analyses. The presented plat<strong>for</strong>m does solve<br />

<strong>the</strong>se problems in <strong>the</strong> domain <strong>of</strong> (mass spectrometry based) proteomics<br />

research. Once <strong>the</strong> raw data is integrated into <strong>the</strong> system each analysis<br />

per<strong>for</strong>med is fully documented and each result is stored and can be<br />

traced back to <strong>the</strong> data and parameter-sets used.<br />

Comparison <strong>of</strong> results: During <strong>the</strong> course <strong>of</strong> a (data centered) project usually<br />

many different algorithms are developed or tested. Additionally,<br />

most algorithms can be parameterized in many ways <strong>of</strong>ten with large<br />

variations in <strong>the</strong> results. The plat<strong>for</strong>m presented here cannot only organize<br />

<strong>the</strong> results depending on algorithms and parameters and thus enable<br />

reasonable comparisons but also provides consistent visualization.<br />

Exploring parameter space: Many algorithms depend on a lot <strong>of</strong> input<br />

parameters. Determining and optimizing <strong>the</strong>se parameters <strong>for</strong> a particular<br />

dataset or certain purpose is not always possible. The presented<br />

plat<strong>for</strong>m in combination with <strong>the</strong> QAD Grid workflow system does not<br />

only allow to create different parameter-sets but also starts <strong>the</strong> desired<br />

algorithms with <strong>the</strong>se input parameters on many different machines in<br />

parallel and monitors <strong>the</strong> execution and can also evaluate <strong>the</strong> results and<br />

send reports to <strong>the</strong> user.<br />

Rapid Grid application development: Thanks to <strong>the</strong> worker concept and<br />

available web-services embedding <strong>of</strong> almost any available program code<br />

becomes possible quite quickly - even on specialized hardware.<br />

6.1 Available Services<br />

The proteomics.net plat<strong>for</strong>m <strong>of</strong>fers <strong>the</strong> following services:<br />

� Web portal <strong>for</strong> (MS-based) proteome research<br />

– Fostering organization, exchange and provision <strong>of</strong> (raw) data and<br />

(intermediate) results)<br />

– Provision and easy usage <strong>of</strong> tools and algorithms (section 6.2.1)<br />

– Providing <strong>of</strong> web services to use available algorithms in local applications<br />

(section 6.2.2)<br />

� Distributed computing (Grid) services<br />

– Enabling processing and analysis <strong>of</strong> proteomics mass data faster<br />

than real time<br />

– Allowing integrating <strong>of</strong> specialized hardware, such as IBM’s Cell/BE<br />

processor in Sony’s Playstation 3 <strong>for</strong> compute intensive tasks<br />

(section 6.2.4)<br />

– Enabling non-Grid applications to use <strong>the</strong> power <strong>of</strong> a distributed<br />

computing environment (section 6.2.5)<br />

– Easy creation and integration <strong>of</strong> Grid applications (section 6.2.6)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!