Using ontologies for preprocessing and mining spectra data on the Grid

Abstract

The analysis of mass spectrometry proteomics data requires the composition of different software tools devoted to the loading, management, preprocessing, mining, and visualization of spectra data. This paper proposes the use of ontologies to guide the composition of preprocessing and data mining tools and describes the approach through MS-Analyzer, a software tool for the integrated management, preprocessing and mining of spectra data on the Grid.

Introduction

Mass Spectrometry (MS) proteomics is a powerful technique for identifying different molecular targets in different pathological conditions [1]. Data produced by MS, the spectra, may be represented as a very large set of measures (intensity, $m/z$), representing the abundance (intensity) of biomolecules having certain mass-to-charge ratio ($m/z$) values. Such data can be used for various analysis, such as biomarker discovery (i.e. a list of peaks characterizing a disease), peptide/protein identification, and sample classification. Since spectra have a high dimensionality and are often affected by errors and noise, preprocessing techniques are required before the application of any data analysis, and especially before Data Mining (DM).

The increasing use of MS in clinical studies causes the collection of spectra data from large sample populations, e.g. to control the progression of a disease. Moreover, the comparative study of a disease may require the analysis of spectra produced in different laboratories, so it is possible to envision that, in a few years, biomedical researchers will need to collect and analyse more and more data coming from remote proteomics laboratories. Grid technology can be useful for providing efficient storage space for maintaining on-line large spectra datasets, the broadband infrastructure needed to collect in a secure and efficient way proteomics data coming from remote laboratories, and the computational power needed by the preprocessing and data mining algorithms.

On the other hand, the building of such a Virtual Proteomics Laboratory involves different technological platforms related to the various steps of a proteomics experiment, such as sample treatments, MS techniques, spectra processing, data mining analysis, and results visualization. Choosing the right tools requires multidisciplinary knowledge from MS specialists to biologists and computer scientists, thus modelling the semantics of processes, tools, and data sources is a key issue for simplifying the design of applications.

This paper proposes the combined use of workflow techniques and domain ontologies as a semantic guide for experiment formulation, tool exploitation, and application design, i.e. services composition. The proposed WekaOntology and ProtOntology ontologies describe concepts and relationships of data mining, bioinformatics software tools, and data sources. The use of such ontologies to compose applications, and the ability to combine in a different way preprocessing and data mining tools, are described by a case study developed with MS-Analyzer.

MS-Analyzer is a Grid-based Problem Solving Environment [5] that uses a Service Oriented Architecture, i.e. it offers spectra analysis services built as a composition of specialized spectra management and data mining services. Functions are thus implemented as Web/Grid Services, i.e. independent, self-describing programs that interact using the following standards: SOAP (Simple Object Access Protocol) for application invocation, WSDL (Web Services Description Language) for service description, and UDDI (Uniform Description, Discovery and Integration) for service publishing and discovery. Using such an approach, MS-Analyzer is able to offer its spectra analysis services to more users, who can submit their analysis tasks in a concurrent way.