Bioinformatics : Gene Expression Data Analysis

Transcription

1 Bioinformatics : Gene Expression Data Analysis Aidong Zhang Professor Computer Science and Engineering

2 What is Bioinformatics Broad Definition The study of how information technologies are used to solve problems in biology Narrow Definition The creation and management of biological databases in support of genomic sequences Oxford English Dictionary (proposed) Conceptualizing biology in terms of molecules and applying information techniques to understand and organize the information associated with these molecules, on a large scale

3 Aims of Bioinformatics Simplest Organize data in a way that allows researchers to access information and submit new entries as they are produced Higher Develop tools and resources that aid in the analysis of data Advanced Use these tools to analyze the data and interpret the results in a biologically meaning manner

4 Subjects of Bioinfromatics Data Source Raw DNA sequence Protein sequence Macromolecular structure Genomes Gene expression Literature Metabolic pathways Data Size 8.2 million sequences (9.5 billion bases) 300,000 sequences (~300 amino acids each) 13,000 structures (~1,000 atomic coordinates each) 40 complete genomes (1.6 million 3 billion bases each) ~20 time point measurements for ~6,000 genes 11 million citations Topics Separating regions Gene product prediction Sequence comparison, alignments, identification Structure prediction, 3D alignment Protein geometry measurements Molecular simulations Phylogenetic analysis Genomic-scale censuses Linkage analysis Clustering, correlating patterns, mapping data to sequence, structural and biochemical data Digital libraries Knowledge databases Pathway simulations

5 Figure taken from

6 DNA Microarray Experiments

7 Gene Expression Data Gene Expression Data Matrix Each row represents a gene G i ; Each column represents an experiment condition S j ; Each cell X ij is a real value representing the gene expression level of gene G i under condition S j ; X ij > 0: over expressed X ij < 0: under expressed A time-series gene expression data matrix typically contains O(10 3 ) genes and O(10) time points.

8 Gene Expression Data sample 1 sample 2 sample 3 genes X 11 X 12 X 13 X 21 X 22 X 23 X 31 X 32 X 33 samples asymmetric dimensionality 10 ~ 100 sample / condition 1000 ~ gene two-way analysis sample space gene space

9 Microarray Data Analysis Analysis from two angles sample as object, gene as attribute gene as object, sample/condition as attribute

10 Challenges of Gene Data Analysis (1) Gene space: Automatically identify clusters of genes which express similar patterns in the data set Robust to huge amount of noise Effective to handle the highly intersected clusters Potential to visualize the clustering results

11 Co-expressed Genes Gene Expression Data Matrix Gene Expression Patterns Co-expressed Genes Why looking for co-expressed genes? Co-expression indicates co-function; Co-expression also indicates co-regulation.

12 Challenges of Gene Data Analysis (2) Sample space: unsupervised sample clustering presents interesting but also very challenging problems The sample space and gene space are of very different dimensionality (10 1 ~ 10 2 samples versus 10 3 ~10 4 genes). High percentage of irrelevant or redundant genes. People usually have little knowledge about how to construct an informative gene space.

13 Sample Clustering Gene expression data clustering

14 Microarray Data Analysis Microaray Data Microarray Images Sample Clusters Gene Expression Matrices Gene Expression Data Analysis Visualization Important Important patterns Important patterns patterns Gene Expression Patterns

15 Our Approaches Density-based approach: recognizes a dense area as a cluster, and organizes the cluster structure of a data set into a hierarchical tree. caculate the density of each data object based on its neighboring data distribution. construct the "attraction" relationship between data objects according to object density. organize the attraction relationship into the "attraction tree". summarize the attraction tree by a hierarchical "density tree". derive clusters from density tree.

16 Our Approaches (2) Interrelated dimensional clustering -- automatically perform two tasks: detection of meaningful sample patterns selection of those significant genes of empirical pattern

17 Our Approaches (3) Visualization tool: offers insightful information Detects the structure of dataset Three Aspects Explorative Confirmative Representative Microarray Analysis Status Numerical methods dominant Visualization serve graphical presentations of major clustering methods Visualization applied Global visualization (TreeView) Sammon s mapping TreeView

18 VizStruct Architecture Explorative Visualization Sample space Confirmative Visualization Gene space

19 VizStruct - Dimension Tour Interactively adjust dimension parameters Manually or automatically May cause false clusters to break Create dynamic visualization

20 Visualized Results for a Time Series Data Set

21 Elements of Clustering Feature Selection. Select properly the features on which clustering is to be performed. Clustering Algorithm. Criteria (e.g. object function) Proximity Measure (e.g. Euclidean distance, Pearson correlation coefficient ) Cluster Validation. The assessment of clustering results. Interpretation of the results.

22 Supervised Analysis Select training samples (hold out ) Sort genes (t-test, ranking ) Select informative genes (top 50 ~ 200) Cluster or classification based on informative genes Class 1 Class 2 g 1 g g 1 g 2... g 4131 g g g

23 Unsupervised Analysis Microarray data analysis methods can be divided into two categories: supervised/unsupervised analysis. We will focus on unsupervised sample classification which assume no membership information being assigned to any sample. Since the initial biological identification of sample classes has been slow, typically evolving through years of hypothesis-driven research, automatically discovering sample pattern presents a significant contribution in microarray data analysis. Unsupervised sample classification is much more complex than supervised manner. Many mature statistic methods such as t-test, Z-score, and Markov filter can not be applied without the phenotypes of samples known in advance.

24 Problem Statement Given a data matrix M in which the number of samples and the volume of genes are in different order of magnitude ( G >> S ) and the number of sample categories K. The goal is to find K mutually exclusive groups of the samples matching their empirical types, thus to discover their meaningful pattern and to find the set of genes which manifests the meaningful pattern.

25 Problem Statement samples Informative Genes gene 1 gene 2 gene 3 gene 4 Noninformative Genes gene 5 gene 6 gene 7 gene 8

26 Problem Statement (2) samples Informative Genes gene 1 gene 2 gene 3 Noninformative Genes gene 4 gene 5 gene 6 gene 7

27 Problem Statement (3) Class 1 Class 2 Class3 Class 1 Class 2 Class3 gene a gene b gene c gene d gene e gene f

28 Related Work New tools using traditional methods : TreeView CLUTO CIT CNIO GeneSpring J-Express CLUSFAVOR SOM K-means Hierarchical clustering Graph based clustering PCA Their similarity measures based on full gene space are interfered by high percentage of noise.

29 Related Work (2) Clustering with feature selection: (CLIFF, leaf ordering, two-way ordering) 1. Filtering the invarient genes Bayes model Rank variance PCA 2. Partition the samples Ncut Min-Max Cut 3. Pruning genes based on the partition Markov blanket filter T-test Leaf ordering

30 Subspace clustering : Bi-clustering d-clustering Related Work (3)

31 Intra-pattern-steadiness Variance of a single gene: Average row variance: = y y S j S i j i y w w S y i Var 2,, ) ( 1 1 ), ( ( ). ) ( 1 1 ), ( 1 ), ( 2,, = = x y y x G i S j S i j i y x G i x w w S G y i Var G y x R We require each genes show either all on or all off within each sample class.

32 Intra-pattern-consistency(2) Measurement residue Data(A) Data(B) MSR ARV*

33 Inter-pattern-divergence In our model, both ``inter-patternsteadiness'' and ``intrapattern-dissimilarity' on the same gene are reflected. Average block distance: D ( x, ( y, y ' )) = i G x w i, S y G x w i, S y '

34 Pattern Quality The purpose of pattern discovery is to identify the empirical pattern where the patterns inside each class are steady and the divergence between each pair of classes is large. Ω = S S 1 R ( x, y ) + R ( x, y, 1 y D ( x, ( y, y )) y 2 )

35 Pattern Quality (2) Data(A) Data(B) Data(C) Con Div W

36 The Problem Input 1. m samples each measured by n-dimensional genes 2. the number of sample categories K Output A K partition of samples (empirical pattern) and a subset of genes (informative space) that the pattern quality of the partition projected on the gene subset reaches the highest.

37 Strategy Starts with a random K-partition of samples and a subset of genes as the candidate of the informative space. Iteratively adjust the partition and the gene set toward the optimal solution. Basic elements: A state: A partition of samples {S 1,S 2, S k } A set of genes G G The corresponding pattern quality W An adjustment For a gene ˇG, insert into G For a gene G, remove from G For a sample in group S, move to other group i g r i g r i s r

38 Strategy (2) Iteratively adjust the partition and the gene set toward the optimal pattern. for each gene, try possible insert/remove for each sample, try best movement.

39 Improvement Data Standardization o the original gene intensity values relative values w ' i, j = w i, j σ i w i, where w i = m m w j= 1 i, j j= 1 ; σi = Random order Conduct negative action with a probability Stimulated annealing m ( w i, j m 1 w ) i 2 p Ω 1 = exp( ) T(0) = 1; T( i) =. Ω T( i) 1+ i

40 Experimental Results Data Sets: Multiple-sclerosis data MS-IFN : 4132 * 28 (14 MS vs. 14 IFN) MS-CON : 4132 * 30 (15 MS vs. 15 Control) Leukemia data 7129 * 38 (27 ALL vs. 11 AML) 7129 * 34 (20 ALL vs. 14 AML) Colon Cancer data 2000 * 62 (22 normal vs. 40 tumor colon tissue) Hereditary breast cancer data 3226 * 22 ( 7 BRCA1, 8 BRCA2, 7 Sporadics)

41 Experimental Results (2) Multiple-sclerosis data CNIO CIT CLUSFAVO R Cluto J-Express Delta EPD* MS_IFN MS_CON

42 Interrelated Dimensional Clustering The approach is applied on classifying multiple-sclerosis patients and IFN-drug treated patients. (A) Shows the original 28 samples' distribution. Each point represents a sample, which is a mapping from the sample's 4132 genes intensity vectors. (B) Shows 28 samples' distribution on 2015 genes. (C) Shows 28 samples' distribution on 312 genes. (D) Shows the same 28 samples distribution after using our approach. We reduce 4132 genes to 96 genes.

43 Experimental Results (3) Experimental Results (3) Leukemia data CNIO CIT CLUSFAV OR Cluto J-Express Delta EPD* G G

44 Experimental Results (4) Experimental Results (4) Colon & Breast data CNIO CIT CLUSFAVO R Cluto J-Express Delta EPD* Colon Brest

45 Applications Gene Function Co-expressed genes in the same cluster tend to share common roles in cellular processes and genes of unrelated sequence but similar function cluster tightly together. Similar tendency was observed in both yeast data and human data. Gene Regulation By searching for common DNA sequences at the promoter regions of genes within the same cluster, regulatory motifs specific to each gene cluster are identified. Cancer Prediction Normal vs. Tumor Tissue Classification Drug Treatment Evaluation

46 Summary We have developed advanced approaches for gene expression data analysis which work more effectively than traditional analysis approaches This research area is exciting and challenging. There are a lot of interesting research issues.