Chalini D Wijetunge1, Isaam Saeed1, Berin A Boughton2, Jeffrey M Spraggins3, Richard M Caprioli4, Antony Bacic5, Ute Roessner2, Saman K Halgamuge1. 1. Department of Mechanical Engineering. 2. Metabolomics Australia, University of Melbourne, Parkville, VIC 3010, Australia. 3. Department of Biochemistry, Mass Spectrometry Research Centre. 4. Department of Biochemistry, Mass Spectrometry Research Centre, Department of Chemistry and Department of Medicine, Vanderbilt University Medical Center, Nashville, TN 37232, USA. 5. Metabolomics Australia, University of Melbourne, Parkville, VIC 3010, Australia, ARC Centre of Excellence, School of Biosciences and Bio21 Molecular Science and Biotechnology Institute, University of Melbourne, Parkville, VIC 3010, Australia.
Abstract
MOTIVATION: Matrix Assisted Laser Desorption Ionization-Imaging Mass Spectrometry (MALDI-IMS) in 'omics' data acquisition generates detailed information about the spatial distribution of molecules in a given biological sample. Various data processing methods have been developed for exploring the resultant high volume data. However, most of these methods process data in the spectral domain and do not make the most of the important spatial information available through this technology. Therefore, we propose a novel streamlined data analysis pipeline specifically developed for MALDI-IMS data utilizing significant spatial information for identifying hidden significant molecular distribution patterns in these complex datasets. METHODS: The proposed unsupervised algorithm uses Sliding Window Normalization (SWN) and a new spatial distribution based peak picking method developed based on Gray level Co-Occurrence (GCO) matrices followed by clustering of biomolecules. We also use gist descriptors and an improved version of GCO matrices to extract features from molecular images and minimum medoid distance to automatically estimate the number of possible groups. RESULTS: We evaluated our algorithm using a new MALDI-IMS metabolomics dataset of a plant (Eucalypt) leaf. The algorithm revealed hidden significant molecular distribution patterns in the dataset, which the current Component Analysis and Segmentation Map based approaches failed to extract. We further demonstrate the performance of our peak picking method over other traditional approaches by using a publicly available MALDI-IMS proteomics dataset of a rat brain. Although SWN did not show any significant improvement as compared with using no normalization, the visual assessment showed an improvement as compared to using the median normalization. AVAILABILITY AND IMPLEMENTATION: The source code and sample data are freely available at http://exims.sourceforge.net/. CONTACT: awgcdw@student.unimelb.edu.au or chalini_w@live.com SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
MOTIVATION: Matrix Assisted Laser Desorption Ionization-Imaging Mass Spectrometry (MALDI-IMS) in 'omics' data acquisition generates detailed information about the spatial distribution of molecules in a given biological sample. Various data processing methods have been developed for exploring the resultant high volume data. However, most of these methods process data in the spectral domain and do not make the most of the important spatial information available through this technology. Therefore, we propose a novel streamlined data analysis pipeline specifically developed for MALDI-IMS data utilizing significant spatial information for identifying hidden significant molecular distribution patterns in these complex datasets. METHODS: The proposed unsupervised algorithm uses Sliding Window Normalization (SWN) and a new spatial distribution based peak picking method developed based on Gray level Co-Occurrence (GCO) matrices followed by clustering of biomolecules. We also use gist descriptors and an improved version of GCO matrices to extract features from molecular images and minimum medoid distance to automatically estimate the number of possible groups. RESULTS: We evaluated our algorithm using a new MALDI-IMS metabolomics dataset of a plant (Eucalypt) leaf. The algorithm revealed hidden significant molecular distribution patterns in the dataset, which the current Component Analysis and Segmentation Map based approaches failed to extract. We further demonstrate the performance of our peak picking method over other traditional approaches by using a publicly available MALDI-IMS proteomics dataset of a rat brain. Although SWN did not show any significant improvement as compared with using no normalization, the visual assessment showed an improvement as compared to using the median normalization. AVAILABILITY AND IMPLEMENTATION: The source code and sample data are freely available at http://exims.sourceforge.net/. CONTACT: awgcdw@student.unimelb.edu.au or chalini_w@live.com SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Authors: Kyle D Bemis; April Harry; Livia S Eberlin; Christina R Ferreira; Stephanie M van de Ven; Parag Mallick; Mark Stolowitz; Olga Vitek Journal: Mol Cell Proteomics Date: 2016-01-21 Impact factor: 5.911
Authors: Patrick M Wehrli; Wojciech Michno; Kaj Blennow; Henrik Zetterberg; Jörg Hanrieder Journal: J Am Soc Mass Spectrom Date: 2019-09-16 Impact factor: 3.109