ClusterSculptor: An Expert Driven Visual Analytics Tool for the Classification of Single Particle Mass Spectra
EMSL Project ID
25639
Abstract
This proposal is a continuation of research under EMSL User Proposal #3644. Single particle mass spectrometers and SPLAT in particular generate massive amounts of complex detailed data. Dr. Zelenyuk's instrument can detect 20 individual particle mass spectra per second, recording a time of detection, size and 450 mass spectral peak intensities for each individual particle. To make efficient use of these multidimensional datasets clearly requires special software. In the past few years we have developed and applied a data classification, visualization and mining software package we call SpectraMiner. Application of this tool to analysis of SPLAT generated data reveals that very often the clustering of particle mass spectra on statistical basis alone produces unsatisfactory results. This is particularly true, when it comes to the classification of mass spectra of particles containing organics. We find that scientists generally have a vast amount of non-compiled knowledge and intuition that they can apply to improve data classification. The idea is to position the expert at the centre of the classification process by giving him/her the power to shape the data in a way that would allow the statistical tools to properly separate and organize the data. We have, in the past year been developing a prototype version of the program we call ClusterSculptor (CS) designed to provide the investigator a visually driven interface of data shaping tools, which is combined with a real-time classification and visualization routines. Preliminary tests suggest that this approach offers enormous promise.
We propose to continue the development of this new approach to data analysis through a collaborative effort between laboratory data generation by SPLAT and software development and testing. The laboratory work will aim to generate datasets of particles composed of atmospherically relevant compounds, generating mass spectral signatures and testing the software capability to identify and classify the mass spectral data. In addition we will be expanding SPLAT's measurement capabilities to include other particle properties, like density, size, and hygroscopicity to improve its quantitative capabilities. In parallel we will expand CS to take these new measures into account during classification.
Project Details
Project type
Large-Scale EMSL Research
Start Date
2007-07-13
End Date
2008-07-13
Status
Closed
Released Data Link
Team
Principal Investigator
Team Members
Related Publications
E. Nam, Y. Han, K. Mueller, A. Zelenyuk, and D. Imre, "ClusterSculptor: A Visual Analytics tool for high-dimensional data," IEEE Symposium on Visual Analytics Science and Technology (VAST), pp. 75-82, Sacramento, CA, November, 2007.
S. Garg, E. Nam, IV. Ramakrishnan, K. Mueller, "Model-Driven Visual Analytics," (to be published, 8 pages), IEEE Symposium on Visual Analytics Science and Technology (VAST), 2008.
Zelenyuk A, DG Imre, EJ Nam, Y Han, and K Mueller. 2008. "ClusterSculptor: Software for Expert-Steered Classification of Single Particle Mass Spectra." International Journal of Mass Spectrometry 275(1-3):1-10. doi:10.1016/j.ijms.2008.04.033