predictPPI: Prediction of the PPI

Description

Major method of the Path2PPI class to predict the final PPI in the target species using the information available from the stored reference species. Different values for the arguments of this method can lead to different PPI networks, differing in the degree of reliability and strictness.

Usage

predictPPI(path2ppi, mode="both", h.thresh=1e-05,  h.range=c(1e-100, 1e-20), i.thresh=0.7,  consider.complexes=FALSE, max.complex.size=5, decline.self.interaction.ref=FALSE, decline.self.interaction.tar=TRUE, verbose=TRUE)

Arguments

path2ppi

An object of the class Path2PPI.

mode

Which interaction from the reference species should be taken into account. "both": both interactors of an interaction has to be in the initial protein list previously inserted by the user (recommended if it is a large network or many proteins were initially defined, respectively). "one": only one of the interactors of each reference interaction has to be in the initial protein list (may lead to very large networks).

h.thresh

E-value cutoff at which each homologous relationship definitely will be declined (see also h.range argument).

h.range

Numeric vector consisting of two values. The first value indicates the lower border (smallest E-value). Each E-value which is equal or less than this border is scored with 1 (best). The second value indicates the upper border (biggest E-value). Each E-value which is equal or greater than this border is scored with 0 (worst).

i.thresh

Numeric. Threshold for accepted interactions. If the computed prediction score for an interaction is less than i.thresh it will be declined.

consider.complexes

Logical. If TRUE then interactions are also considered which actually indicate an association of the current protein to one bigger protein complex. This may lead to very large networks if mode="one" since all other proteins of this complex are considered as well, i.e., each protein in such complexes are considered to interact with each other protein of this complex. If mode="both" then each protein of an complex has to be in the initial protein list to consider each interaction (see details).

max.complex.size

Numeric. Is only considered if consider.complexes=TRUE. The maximum size of complexes to be considered.

decline.self.interaction.ref

Logical. If TRUE then all self interactions from reference species are declined.

decline.self.interaction.tar

Logical. If TRUE then all predicted self interactions in target species are declined.

verbose

Logical. FALSE hides messages in the output.

Value

An object of the class Path2PPI with predicted PPI.

Details

Difference of h.thresh and h.range: If only one protein in the target species was found to be homologous to a current reference species protein and this homology was rated with an E-value which is equal or smaller than h.thresh it is scored with 1 (even if the E-value is larger than the upper border of h.range). See package vignette for more details.

Use the complex arguments with care, since each complex may lead to a vast amount of interactions, i.e., each protein is considered to interact with each other of this complex; e.g. if there are 10 proteins involved in one complex, this would lead to 10 over 2 = 45 interactions.

Examples

Run this code

data(ai) #Load test data set

ppi <- Path2PPI("Autophagy induction", "Podospora anserina", "5145")
ppi <- addReference(ppi, "Homo sapiens", "9606", human.ai.proteins, 
                    human.ai.irefindex, pa2human.ai.homologs)
ppi <- addReference(ppi, "Saccharomyces cerevisiae (S288c)", "559292", 
                    yeast.ai.proteins, yeast.ai.irefindex, 
                    pa2yeast.ai.homologs)

#Using the default settings leads to 8 predicted interactions in the 
#target species
ppi <- predictPPI(ppi)

#Consider complexes where each complex is allowed to be up to 10 proteins 
#large. For this smaller pathway only one more interaction was predicted when 
#considering larger complexes.
ppi <- predictPPI(ppi,consider.complexes=TRUE,max.complex.size=10)

#We can be less strict and decrease h.range what obviously increases the 
#number of predicted interactions to 13
ppi <- predictPPI(ppi,h.range=c(1e-60,1e-20))