d. Return D3 and replace the elements that exist in D and D3 with random numbers. Until Initially, our algorithm removes genes or conditions VEGFR or time points from the dataset to accomplish largest dimin ishing of score S this Inhibitors,Modulators,Libraries step is described in the following section in which a node corresponds to a gene or experi mental condition or time point in the 3D microarray gene expression dataset. Algorithm Inhibitors,Modulators,Libraries 2 Input. D, a matrix of real numbers that represents 3D Delete gene or sampleexperimental condition or time point that has highest u score and modify I or J or K. Recalculate miJK, i I mIjK, j J mIJk, k K mIJK and S. End while Return M The complexity of ?rst and second steps is O as those will iterate times. The complexity of selec tion of best genes, samples and time points is O.

So it is suggested to use algorithm Inhibitors,Modulators,Libraries II before algorithm 3. As the goal of our algorithm is to ?nd maximal triclusters, having MSR score below Inhibitors,Modulators,Libraries the thresh old, the resultant tricluster M may not be the largest one. That means some genes andor experimental End if Until The complexity of this algorithm is O where m, n and p are the number of genes, samples and time points in the 3D microarray dataset. In the second step, we delete one node at each iteration from the resultant submatrix, produced by Algorithm 2, until the score S of the resultant submatrix is less than or equal to. This step results in a tricluster. Algorithm 3 Input. D, a matrix of real numbers that represents 3D microarray gene expression dataset 0, maximum allowable MSR threshold. Output.

MIJK, a tricluster, consisting of a subset of genes, a subset of samplesexperimental Inhibitors,Modulators,Libraries conditions conditionssamples andor time points may be added to the resultant tricluster T produced by node deletion algo rithm, so that the MSR score of new tricluster T produced after node addition does not exceed the MSR score of T. Now the third step of our algorithm is described below. Algorithm 4 Input. D, a matrix of real numbers that represents tricluster, having a subset of genes, a subset of experimental conditionssamples and a subset of time points. Output. MI J K, a tricluster, consisting of a subset of genes, a subset of samplesexperimental conditions and a subset of time points, such that II, J J, K K and MSR MSR of D. Initialization.

M D Recalculate mIjK, j mIJk, mIJK and S Add samplesexperimental conditions j J that satisfy the following inequality Recalculate miJK, i mIJk, k mIJK and S Add time points k K that satisfy the following inequality Tricluster eigengene To ?nd tricluster eigengene we applied singular value decomposition method on the expression data of each tricluster. For instance, Xg represents enzyme inhibitor the expression matrix of ith tricluster, where g, c and t rep resent the number of genes, samples and time points of ith tricluster. Now we apply SVD on the data matrix. Now, the SVD of ith tricluster can be represented as, where U and V are the orthogonal matrices.