TANG16 / MMI

Codes for Multimodel mutual information

Six scripts to calculate MMI (Multimodal Mutual Information) and MDI (Multimodal Direct Information)

1. dynamicprogramming.m is used for clustering the expression profiles for each gene into clusters by dynamic programming.

Parameters: a. Data: gene expression data. each row represents one sample, each column for each gene. b. maxallow: the maximum cluster number allowed. The program calculates the BIC values from 1 to maxallow to choose the exact cluster number with the smallest BIC. Values:

2. MMI.m is used to caluculate multimodal mutual information for gene pairs.

Parameters: a. Both X and Y are column vectors, containing the gene expression profiles for one gene, respectively. b. Both clusterlabelX and clusterlabelY are column vectors, they store the clustering results for gene X and Y.

3. MDIin.m is used to caluculate the covariance and inverse covariance matrix of inner MDI.

Parameters: a. expressdata: gene expression data. each row represents one sample, each column for each gene. b. clusterassign: the clustering result for all the genes. Each row is a sample, each column is one gene. c. clusternum: the number of cluster for all the genes. d. shrink: shrinkage parameter for inversion.

4. MDIcombine is used to calculate multimodal direct information.

Parameters: a. data: gene expression data. each row represents one sample, each column for each gene. b. clusterassign: the clustering result for all the genes. Each row is a sample, each column is one gene. c. clusternum: the number of cluster for all the genes. d. MDIin_invMatrix: inverse covariance matrix of inner MDI. The output of MDIin.m e. shrink: shrinkage parameter for matrix inversion.

5. main.m is used to test the scripts of MDI and MMI.

6. myfuncov.m is used to approximate the real covariace matrix by constrain optimization.