Page 1 |
Save page Remove page | Previous | 1 of 200 | Next |
|
small (250x250 max)
medium (500x500 max)
Large (1000x1000 max)
Extra Large
large ( > 500x500)
Full Resolution
All (PDF)
|
This page
All
|
SPARSE REPRESENTATION MODELS AND
APPLICATIONS TO BIOINFORMATICS
by
Roger Pique-Regi
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ELECTRICAL ENGINEERING)
August 2009
Copyright 2009 Roger Pique-Regi
Object Description
| Title | Sparse representation models and applications to bioinformatics |
| Author | Pique-Regi, Roger |
| Author email | rpique@gmail.com; piquereg@usc.edu |
| Degree | Doctor of Philosophy |
| Document type | Dissertation |
| Degree program | Electrical Engineering |
| School | Viterbi School of Engineering |
| Date defended/completed | 2009-01-30 |
| Date submitted | 2009 |
| Restricted until | Unrestricted |
| Date published | 2009-08-05 |
| Advisor (committee chair) | Ortega, Antonio |
| Advisor (committee member) |
Kosko, Bart Asgharzadeh, Shahab |
| Abstract | Microarrays and new sequencing techniques offer a high throughput platform to study the whole genome with the unprecedented capability of measuring millions of genomic features on a single essay. This massive parallel measurement power has an enormous potential for research in Biology and Medicine with the ultimate objective of identifying and learning the biological processes occurring in different organisms and diseases. Existing model learning methods are severely limited by the reduced number of samples that are usually available compared to the measurements.; We propose that sparse signal representations can model these biological signals and we develop the analytical tools to accurately extract the relevant information. We exploit the underlying sparseness of the biological model to overcome some of the problems associated with analyzing these massive datasets. This work demonstrates the potential benefits of this approach by studying three different problems involving microarray data.; The first problem concerns the supervised design with a limited amount of training samples of a classification procedure to predict tumor progression. We propose a greedy search strategy to select a sparse feature subset with a block diagonal covariance matrix structure to build a linear discriminant analysis model for tumor prognosis. The second problem deals with the detection of copy number alterations. We develop a maximally sparse representation for these copy number alterations, and a sparse Bayesian learning approach is optimized to detect these alterations from noisy microarray observations. The third problem involves finding genetic determinants of gene expression. In this case, we propose a linear regression model with a sparse Bayesian prior on the large matrix of the regression coefficients relating genome alterations to transcription levels. |
| Keyword | bioinformatics; copy number; linear discriminant analysis; sparse Bayesian learning; sparse models; sparse representations |
| Language | English |
| Part of collection | University of Southern California dissertations and theses |
| Publisher (of the original version) | University of Southern California |
| Place of publication (of the original version) | Los Angeles, California |
| Publisher (of the digital version) | University of Southern California. Libraries |
| Provenance | Electronically uploaded by the author |
| Type | texts |
| Legacy record ID | usctheses-m2484 |
| Contributing entity | University of Southern California |
| Rights | Pique-Regi, Roger |
| Repository name | Libraries, University of Southern California |
| Repository address | Los Angeles, California |
| Repository email | cisadmin@dots.usc.edu |
| Filename | etd-Regi-2698 |
| Archival file | uscthesesreloadpub_Volume26/etd-Regi-2698.pdf |
Description
| Title | Page 1 |
| Contributing entity | University of Southern California |
| Repository email | cisadmin@dots.usc.edu |
| Full text | SPARSE REPRESENTATION MODELS AND APPLICATIONS TO BIOINFORMATICS by Roger Pique-Regi A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) August 2009 Copyright 2009 Roger Pique-Regi |
