Gibbs sampling for microarray data analysis
1. Introduction
1.1 Bioinformatics: a general introduction on its origin and sub-disciplines
1.2 Microarray: the technology, its growing importance in bioinformatics, and its characteristics 1.3 Clustering algorithms: the introduction of traditional clustering methods to bioinformatics, the
limitation of the traditional methods, and an overview of popular clustering algorithms applied for microarray data analysis
1.4 Biclusterin algorithms: the different biclustering problems on microarray data, and the different strategies.
1.5 Gibbs sampling: the basic idea, the introduction of Gibbs sampling method to bioinformatics (i.e. its application to motif-finding problems), why Gibbs sampling for the biclustering problem
2. Bayesian models for microarray data analysis 2.1 Characteristics of microarray data (in more detail)
2.2 Basic ideas of Bayesian methods: posterior likelihood * prior 2.3 Applying Bayesian models on microarray data
2.3.1 Advantages
2.3.2 Which distributions to choose: taking into the account of the noisy characteristics of microarray data
2.3.3 The preprocessing of microarray data: (in order to apply the specified models)
3. Biclustering for microarray data:
3.1 For pathologica discovery: finding subsets of patients who have coherent gene expressions over a subset of genes:
3.1.1 The discrete model
3.1.2 The Gibbs sampling strategy 3.1.3 Results
3.2 For discoveries in systems biology:
3.2.1 The continuous model 3.2.2 Model for time-sequence data 3.2.3 The Gibbs sampling strategy 3.2.4 Results
4. Query-driven biclustering on microarray data:
4.1 The introduction of prior knowledge 4.2 Applications in pathological discovery
4.2.1 The modified model and the Gibbs sampling procedure 4.2.2 Results
4.3 Applications in systems biology
4.3.1 The modified model and the Gibbs sampling procedure 4.3.2 Results
5. Comparison of the Gibbs biclustering strategy to other methods:
5.1 Comparison between Bayesian and non-Bayesian methods 5.2 Parameterization: Gibbs sampling vs. EM
6. Future work:
6.1 Incorporating data of different sources into the Bayesian model and the Gibbs sampling scheme