WSU BiRG

  • Increase font size
  • Default font size
  • Decrease font size
Home
A web-based framework for the distribution of bioinformatics techniques: Orthogonal projection on latent structures and principal component analysis implemented as RESTful web services
Research Area: Bioinformatics Year: 2009
Type of Publication: In Proceedings  
Authors:  
Book title: 2009 Ohio Collaborative Conference on Bioinformatics
Month: June
   
Note:
Abstract and poster, Case Western Reserve University, Cleveland, Ohio
Abstract:
While bioinformatics has changed the study of biology in many respects, realization of its full potential has in many instances been hindered by the separation of algorithm developers and laboratory scientists. To bridge this gap, algorithm developers provide their source code and in some instances, precompiled versions for specific platforms. Even with these resources available, there are often additional complications (e.g., missing dependencies). These shortcomings can be overcome through the adaption of a web-based framework, where novel algorithms are available online for scientists to immediately incorporate into their data analysis. This poster demonstrates such a framework by implementing two common techniques for analyzing “omics” data: Orthogonal Projection onto Latent Structures (OPLS) and Principal Component Analysis (PCA). These two techniques are implemented in a web-based framework that allows interaction through a web browser or as a RESTful web service. The RESTful services paradigm is an implementation of service–oriented architecture, based on Representational State Transfer (REST). REST advocates a resource first perspective of a distributed system, in which each of the resources can be manipulated using a set of simple commands. This simplicity eliminates many of the layers found in the alternative SOAP approach for services. By providing a web service interface for each technique (OPLS and PCA), researchers can incorporate the techniques into scientific workflows or develop their own customized programs that leverage existing techniques. Further, by also providing a standalone web interface for each technique, both OPLS and PCA can be effectively applied to experimental data without compiling source code or installing any additional dependencies. Such techniques will be demonstrated using a NMR-based metabolomics dataset from a rat model of tissue toxicity.