Fast parallel multivariate analysis of mass spectrometry imaging data: A new approach for studying human liver disease

Project completed 2014.

Dr Josephine Bunch, School of Chemistry
Dr Hamid Dehghani, School of Computer Science
Dr Patricia Lalor, Institute of Immunology and Immunotherapy
Dr Iain Styles, School of Computer Science
Dr Ata Kaban, School of Computer Science

In mass spectrometry imaging, measurements which contain information about the masses of the chemical constituents of the sample are taken from many different locations. These measurements can be formed into a "mass image" which provides extraordinary detail about the spatial variations of the chemical composition of the sample. For example, it is possible to examine where a particular lipid or a drug is localised within a particular type of tissue. In a typical experiment measurements may be taken from ~10,000 points; each of which consists of the abundance of chemical components at >100,000 different mass-to-charge ratios. The principled interpretation of so much data is extraordinarily challenging.

We propose to develop new analysis strategies for these large multivariate datasets that take advantage of the latest developments in computing technology, notably the large-scale and easy availability of graphics processing units, which will enable the large-scale parallelisation of the data processing. We will use these new technologies to develop fundamentally new algorithms for the analysis of this data that will enable it to be done rapidly and in a memory-efficient manner (many datasets are too large to fit in the memory of most computers). We will also develop techniques inspired by recent advances in computer vision that will allow patterns to be extracted in a hierarchical way that should provide new insight into the structures and relationships that are embedded deep within these enormous datasets. These advances have significant potential benefits in several areas of biomedical research in which we will seek to apply these methods, including drug discovery, liver disease, and trauma research.

Link to ethesis: