Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.583948
Title: Signal processing techniques for the interpretation of microarray measurements
Author: Bowles, Thomas
Awarding Body: Cardiff University
Current Institution: Cardiff University
Date of Award: 2006
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Microarray technology allows the measurement of gene transcription on a genome wide scale. Signal processing approaches to the analysis of data from microarray time course experiments are the focus of this thesis. Firstly, spectral estimation methods are explored as a method for the detection of cell-cyclic elements within microarray data. High resolution data-dependent filterbank methods are proposed as an improvement to the traditional periodogram approach. A spectral estimator is then designed specifically to deal with the errors in the sampling times inherent in microarray experiments, which is based on the robust Capon beamformer. A beamforming inspired approach is shown to yield a more robust, and higher resolution, estimate of the magnitude spectrum of the whole data set than the previous spectral estimation approaches. Blind source separation is examined as a method for recovering sources which represent fundamental cellular processes. The linear mixing model is compared to its transpose form, and a dual form, in terms of their finite sample performance with real microarray data. Second order methods are proposed to recover sources which are spatio-temporally uncorrelated and may be more suitable with microarray data. Both the spectral and blind source separation techniques are shown to yield useful feature extraction measures for microarray data clustering. The spectral feature extraction allows the clustering of cell-cyclic genes into a single functional group. Finally, sparse source separation is introduced as a possible blind separation technique with microarray data.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.583948  DOI: Not available
Share: