Use this URL to cite or link to this record in EThOS:
Title: Bayesian biclustering algorithms and their application to gene expression data
Author: Fowler, Anna
ISNI:       0000 0004 2752 7152
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Access from Institution:
Biclustering identifies subsets of rows and columns of a data matrix which are similar, revealing an underlying structure which may be obscured when clustering either rows or columns independently. Gene expression data sets typically contain measurements from thousands of genes for a comparatively small number of samples, or time points. Therefore, biclustering models and algorithms need to be robust to imbalanced dimensions if they are to be effectively applied to gene expression data. In this thesis, we develop biclustering algorithms for application to gene expression data. A novel, agglomerative biclustering algorithm for application to a data matrix containing observations from a far larger number of genes than samples is introduced. Variable selection is included in this algorithm, allowing sample clusters to be based on only the discriminating gene clusters. This algorithm is developed to ensure a fast computation time and the effect of having a far larger number of genes than samples is examined. Dynamic clusters which split and merge over time are introduced for time series data; these allow cluster memberships to be time dependent, forming biclusters of genes or samples and time points. Reversible Jump MCMC methods are used to identify these clusters, and alternative proposal schemes are considered to improve acceptance rates and overall performance of the algorithm. Combining these two ideas produces biclusters of genes and samples which split and merge over time, identifying co-regulated subsets of genes, samples and time points.
Supervisor: Heard, Nicholas ; Adams, Niall Sponsor: Engineering and Physical Sciences Research Council
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available