ScISI (version 1.44.0)

maximizeSimilarity: A function compares two bipartite graph matrices and finds the most similar matches between the clusters

Description

This function takes a matrix of similarity indices between two bipartite graph matrices and determines, for each complex of the first bipartite graph matrix (bg1), the most similiar complexes of the second bipartite graph matrix (bg2).

Usage

maximizeSimilarity(simMat, bywhich = "ROW", zeroSim = "NO")

Arguments

simMat

A matrix of similarity coefficients between two bipartite graph matrices (bg1 and bg2) where the rows are indexed by the first bipartite graph matrix and the colunms are indexed by the second bipartite graph matrix. The (i,j)th entry is the similarity index between complex i of bg1 and complex j of bg2.

bywhich

Takes one of these three arguments: "ROW", "COL", "BOTH"

zeroSim

Takes either one of the following arguments: "NO", "YES"

Value

The return value is a list consisting of a vector and a list:

maximize

A named numeric vector. The name is the complex, C-i, for which the function is trying to find a maximal match. The entries of the vector is the maximal similarity index between C-i and all of the complexes of the other bipartite graph matrix, i.e. the maximal entry row i in simMat.

maxComp

A named list of named vectors. The name is the complex, C-i, for which the function is trying to find a maximal match. The named vector consists of the positions of the maximal matches (either which row or which column) and the names correspond to the conmplex of maximal matching.

Details

This function's purpose is to take one (or both) bipartite graph matrix, wlog we take bg1, and, for each complex, C-i, of bg1, finds the complex(es) of bg2 that are the most similiar to C-i based on the similarity index. Since the complexes of bg1 is indexed by the rows of simMat argument, finding the complexes of bg2 that are the most similar to C-i means finding the maximal value, m, of row i and then the complexes, K-j, that index the colunm for which m belongs.

If byWhich argument is set to "ROW", the function parses through each complex of bg1 and finds the complex(es) of bg2 which are most similar. If byWhich is set to "COL", the function parses through each complex of bg2 and finds the complex(es) of bg1 which are most similar. If byWhich is set to "BOTH", the function parses through both the complexes of bg1 and bg2. Since this matrix is not symmetric (this matrix is usually not square) this maximizing is different between row and column.

If zeroSim argument is set to "NO", the only maximal matching occurs if the similarity index is nonzero; e.g. if we want to maximize the match for complex C-i of bg1, but row i is comprised only of 0, C-i will not be matched to any complex of bg2.

Examples

#go = getGOInfo(wantAllComplexes = FALSE)
#mips = getMipsInfo(wantSubComplexes = FALSE)
#goM = createGOMatrix(go)
#mipsM = createMipsMatrix(mips)
#cc = runCompareComplex(mipsM, goM, byWhich="ROW")
#m = maximizeSimilarity(cc$JC, byWhich = "ROW")
#m$maximize
#m$maxComp