During the ASMS Sanibel conference on MSI in January 2007 a discussion began concerning the merits of a database of sample preparation protocols to enable MSI best-practice and aid new researchers entering the field. The problem of a lack of standardization and training in MSI was compellingly confirmed during the first NordForsk funded MSI training course (organized by Corthals, McDonnell and Heeren and held at the FOM Institute AMOLF in March 2009): despite most participants using the same commercial matrix deposition device its practical usage differed widely (and in most cases contrary to the manufacturer’s recommendations). With suitable training all participants were able to generate near identical, high quality MSI datasets.
A similar situation was evident in the subsequent data analysis training course (organized by Corthals and McDonnell and NordForsk funded, at Turku Centre for Biotechnology in December 2009). The majority of participants were restricted to the data analysis tools included in the software provided with their commercial instruments or the freely available Biomap software, an independent package developed for MSI by Markus Stoeckli (and which has been an essential element in its rapid uptake and development). Nevertheless the single biggest impediment to effective data analysis was a lack of training to fully utilize the statistical methods that were available. Following the training course the participants were able to generate robust classifiers and perform molecular histology much more effectively.
The restriction to data analysis packages supplied with the commercial mass spectrometers or Biomap means that most researchers have not been able to exploit the improved data analysis capabilities reported by data analysis specialists. A lack of a common data format standard is one significant reason most newly reported capabilities have not been utilized as much as they may otherwise be; an open-source MSI data analysis platform that incorporates the latest algorithms would certainly help but would only be part of the answer: the researchers must understand the data analysis algorithms in order to use them correctly. The lengthy introduction to MSI data analysis reported by Jones et al. was written to begin to address this need.
Only with sufficient training and co-operation can the full potential of MSI be utilized to test the capabilities of these highly cross-disciplinary tools against an array of diseases of present day concern, both in terms of improved diagnosis and pharmacological development. Central to this purpose will be the dissemination of the complementary techniques and expertise developed in independent research laboratories and their sustained interaction. Interaction between MSI researchers is crucial for devising best-practice guidelines and web-based experimental resources; the involvement of healthcare researchers is essential to ensure MSI targets real needs in healthcare research and pharmaceutical development.
In 2010 an application was made for a COST Action research network, to explicitly fund the cooperation needed to devise best practice guidelines in MSI, identify synergies in current MSI methods, and improve the accessibility of the technique by providing detailed training courses and opportunities for short term placements in Europe’s leading MSI laboratories. The first application was unsuccessful, falling at the final fence, but encouragingly we were invited to reapply the next year. The 2011 application was successful; one of just four COST Actions awarded out of more than one hundred applications, and will now run until November 2015.
COST Action BM1104, entitled “Mass Spectrometry Imaging: New Tools for Healthcare Research” involves all major European pharmaceutical companies and is supported by the MS vendors and the European Proteomics Association, which has made MSI one of four special initiatives. The central idea behind the COST Action is information exchange and training; for data acquisition, data analysis, and their application and to provide this knowledge as a public resource. The COST Action proposal (the Memorandum of Understanding) can be downloaded here.
Data Acquisition – Best Practice Guidelines
Imaging MS requires the localized extraction of the molecules of interest followed by spatially correlated mass analysis. The sample preparation and mass analysis methods are critical factors that determine which molecules are measured, and the sensitivity and resolution at which they can be detected. Imaging MS and healthcare researchers will visit each other’s laboratories to test the performance of the imaging MS methods (sample preparation, mass analysis) that have been developed in each laboratory. The explicit inclusion of multiple pathologies and multiple practitioners provides the capacity and redundancy to begin devising best practice guidelines for multiple molecular classes and tissues.
Histology-defined analysis can be used for the identification of biomarkers specific to pathological entities, and histology-independent analyses examine and classify the tissue solely on the basis of their MS signatures. Both of these approaches have the potential to generate new diagnostic tools and many data analysis techniques have been developed. However as most imaging MS experiments have been performed using commercial instruments using proprietary data formats, many clinical users have been ‘locked’ into single data analysis packages and have not been able to exploit the new data analysis capabilities. A new imaging data standard, imzML, was developed within the 6th framework program Computis. Substantial support from instrument vendors has led to imzML being implemented as an export option on most commercially available instruments. The different data analysis capabilities developed in the partner laboratories will be made imzML compliant to enable widespread data sharing and an explicit comparison of the different imaging data analysis tools. Comparing the performance of the data analysis tools for a variety of pathologies will establish standardized tools and context-dependent best-practice guidelines for analyzing such rich datasets.