Theory that expression of interacting entities is correlated due to evolutional or physical reasons makes it possible to predict networks of interactions from expression values. This is a good opportunity to start with, especially if you have no initial hypothesis concerning your gene expression data.
8.1 Color network by expression values.
To overlay Expression Experiment results onto an existing pathway diagram:
Fig.8.1 Color network by expression values.
8.2 Extract pathways from expression data.
Correlation algorithms group genes according to similarities in patterns of expression variation over all the samples. A correlation network is a group of genes whose expression profiles are highly predictive of one another. Each pair of genes related by a correlation coefficient larger than a minimum threshold and smaller than a maximum threshold (assigned in the initialization dialog box) is connected by a edge. Groups of genes connected to one another are referred to as networks.
Parameters:Sample Selection
The sample selection option indicates whether to cluster genes or samples.
Use Permutation Test
This check box is used to indicate that the minimum threshold R2 value should be selected based on a distribution constructed from element to element R2 values derived following permutation of the expression vectors.
Min Threshold
This value ranging from 0 to 1.0 indicates the smallest R2 possible between two elements to permit a link between the elements in a subnet.
This is minimal correlation which you want to include in your network.
Max Threshold
This value ranging from 0 to 1.0 indicates the greatest R2 possible between two elements to permit a link between the elements in a subnet.
This is maximal correlation which you want to include in your network.
Use Filter
This option allows the user to filter out elements with little dynamic change thus removing flat or uninteresting elements. A measure of entropy is used to rank the elements. The percentage value entered (1 to 100) indicates what percentage of the elements to retain for the construction of the network. A value of 25 will retain the 25% of elements having the greatest entropy.
Distance Metric: Pearson squared
Fig.8.2 Extract pathways from expression data.
The algorithm calculates the correlation coefficient between genes by comparing the expression pattern of each gene to that of every other gene. The ability of each gene to predict the expression of each other gene is measured as a correlation coefficient. Genes are represented as nodes in a network and edges are drawn between them if their correlation coefficient falls between the minimum and maximum thresholds specified in the initialization dialog. The experiment subtree created in the Project Properties Panel contains information regarding the networks predicted. Under the Network tab is a graph of all of the subnets generated. A subnet is a group of genes in which each gene is connected to at least one other gene. The Correlation Subnets tab contains network diagrams for each of the individual subnets, and the Expression Images folder contains expression views for the genes in each of them.
8.3 Build Pathways for selected expression values
To create a new pathway from an expression experiment:
To create a new group from an expression experiment, follow these steps:
Fig.8.3 Build Pathways for selected expression values.