RANKS (version 1.1)

weighted.score.single.vertex-methods: Single vertex score functions - weighted version

Description

Methods to compute weighted score functions applied to a single vertex of the graph

Usage

# S4 method for graph
single.NN.w.score(RW, x, x.pos, w)
# S4 method for matrix
single.NN.w.score(RW, x, x.pos, w)
# S4 method for graph
single.KNN.w.score(RW, x, x.pos, w, k = 3)
# S4 method for matrix
single.KNN.w.score(RW, x, x.pos, w, k = 3)
# S4 method for graph
single.eav.w.score(RW, x, x.pos, w, auto = FALSE)
# S4 method for matrix
single.eav.w.score(RW, x, x.pos, w, auto = FALSE)

Value

single.NN.w.score: the weighted NN score of the vertex

single.KNN.w.score: the weighted KNN score of the vertex

single.eav.w.score: the weighted Empirical Average score of the vertex

Arguments

RW

matrix. It must be a kernel matrix or a symmetric matrix expressing the similarity between nodes

x

integer. Index corresponding to the element of the RW matrix for which the score must be computed

x.pos

vector of integer. Indices of the positive elements of the RW matrix

w

vector of numeric. Its elements represent the initial likelihood that the nodes of the graph belong to the class under study. The elements of w correspond to the columns of RW and the length of w and the number of columns of RW must be equal.

k

integer. Number of the k nearest neighbours to be considered

auto

boolean. If TRUE the components $K(x,x)+K(x_i,x_i)$ are computed, otherwise are discarded (default)

Methods

signature(RW = "graph")

single.NN.w.score computes the weighted NN score for a single vertex using a graph of class graph (hence including objects of class graphAM and graphNEL from the package graph)

single.KNN.w.score computes the weighted KNN score for a single vertex using a graph of class graph (hence including objects of class graphAM and graphNEL from the package graph)

single.eav.w.score computes the weighted Empirical Average score for a single vertex using a graph of class graph (hence including objects of class graphAM and graphNEL from the package graph)

signature(RW = "matrix")

single.NN.w.score computes the weighted NN score for a single vertex using a kernel matrix or a symmetric matrix expressing the similarity between nodes

single.KNN.w.score computes the weighted KNN score for a single vertex using a kernel matrix or a symmetric matrix expressing the similarity between nodes

single.eav.score computes the weighted Empirical Average score using a kernel matrix or a symmetric matrix expressing the similarity between nodes

Details

single.NN.w.score computes the weighted NN score for a single vertex:

$$score(x)=-\underset{x_i\in V_C}{min}-2K(x,x_i))\ast w(x_i)$$

where $V_C$ is the set of positive vertices, and $w(x_i)$ is the weight associated to the node $x_i$

single.KNN.w.score compute the weighted KNN score for a single vertex:

$$score(x)=\sum _{knearest{x}_i\in V_C}2K(x,x_i)\ast w(x_i)$$

single.eav.score computes the weighted Empirical Average score for a single vertex:

$$score(x)=-K(x,x)\ast w(x)+\frac{2}{(\sum _{x_i\in x.pos}w(x_i))}\ast \sum _{x_i\in x.pos}K(x,x_i)\ast w(x_i)$$

See Also

Methods for scoring a single vertex Methods for scoring multiple vertices - weighted version

Examples

# Computation of scores using STRING data with respect to 
# the FunCat category 11.02.01 rRNA synthesis 
library(bionetdata);
data(Yeast.STRING.data);
data(Yeast.STRING.FunCat);
labels <- Yeast.STRING.FunCat[,"11.02.01"];
n <- length(labels);
ind.pos <- which(labels==1);
# NN-score  computed directly on the STRING matrix on the first yeast gene YJR121W
s <- single.NN.w.score(Yeast.STRING.data, 1, ind.pos, w=labels);
# NN-score weighted computed directly on the STRING matrix on the first yeast gene YJR121W,
# using this time random weights for the value of positive nodes
w <- runif(n);
s <- single.NN.w.score(Yeast.STRING.data, 1, ind.pos, w=w);
# \donttest{
# NN-score weighted computed on the 1 step and 2-step random walk kernel matrix
K <- rw.kernel(Yeast.STRING.data);
sK <- single.NN.w.score(K, 1, ind.pos, w);
K2 <- p.step.rw.kernel(K, p=2);
sK2 <- single.NN.w.score(K2, 1, ind.pos, w);
# }