LncMod (version 1.1)

tri.app: Modulators Prediction

Description

This function predicts a modulator affecting the ability of a effector to regulate its targets based on expression profiles.

Usage

tri.app(ms, ET, M.exp, E.exp, T.exp, N = 0.25, method = "pearson", iqr.filter = c(log2(1.5), log2(1.5), log2(1.5)),  cor.MvsET = c(0.3, 0.3), cor.EvsT.dif = 0.45, cor.EvsT.abs = 0.4,  ET.fc.filter = log2(1.5) , ET.p.filter = 0.01 ,  rand = 100,correction="BH",cores=1)

Arguments

ms

a character string (vector) specifying candidate modulator names to predict.

ET

a dataframe representing effector-target regulations in which factors are effector/target names.

M.exp

a numeric dataframe representing expression profile of candidate modulator whose rownames is the candidate modulator names.

E.exp

a numeric dataframe representing expression profile of effectors whose rownames are effector names and column number must be equal to M_exp.

T.exp

a numeric dataframe representing expression profile of targets whose rownames are target names and column number must be equal to M_exp.

N

a numeric (rangs from 0 to 0.5,default 0.25) specifying proportion,by which LOW/HIGH sample group are extracted from a sorted modulator expression samples.

method

a character string (default "pearson") indicating which correlation coefficient is to be computed. One of "pearson" (default), "kendall", or "spearman", can be abbreviated.

iqr.filter

a numeric vector of the form c(modulator_iqr,effector_iqr,target_iqr) specifying IQR threshold to filter expression profiles (default (log2(1.5), log2(1.5), log2(1.5))).

cor.MvsET

a numeric vector of the form c(cor.MvsE,corMvsT) specifying threshold for correlation between modulator and effector/target (default (0.3, 0.3)).

cor.EvsT.dif

a numeric (default 0.45) specifying threshold for difference between effector-target correlation in LOW and HIGH sample group.

cor.EvsT.abs

a numeric (default 0.4) specifying threshold for effector-target correlation either in LOW sample group or HIGH.

ET.fc.filter

a numeric specifying threshold for fold change of effector expression in LOW sample group versus HIGH (default log2(1.5)).

ET.p.filter

a numeric specifying threshold for p value of effector expression in LOW sample group versus HIGH (default 0.01).

rand

a numeric specifying the number of disturbance (default 100).

correction

Correction method (default "BH") in one of p.adjust.methods.

cores

The number of cores (default 1) to use, i.e. at most how many child processes will be run simultaneously. Must be exactly 1 on Windows (which uses the master process).

Value

A list containing following components:

• triplets predicted triplets and related information,a 7 columns dataframe as following:
  • modulator effector target represented modulator/effector/target names,respectively;
  • R_low R_high effector-target correlation in LOW/HIGH sample group,respectively;
  • P_value significance of the triplet;
  • fdr corrected P_value by the assigned method;
• initialnot names of modulators whose expression is not in initial expression profile (M.exp);
• filterdnot names of modulators whose expression is in initial expression profile but not in filterd profile by IQR;

Details

Note:All the arguments without default value must be assigned.

This function running a time checked whether a modulator in a sets,one by one,can affect the ability of a effector sets to regulate their corresponding targets.Please go to Kai Wang,M. et al. Genome-wide identification of post-translational modulators of transcription factor activity in human B cells. Nature biotechnology,27, 829-837 (2009) for detailed information.

The running time and the memory required was increasing as the possible triplets increased.To speed-up the analysis,the function implemented parallel computations when runned on multi-core machines. It used mclapply function in the parallel package to make use of all the CPUs available on the system, with each core simultaneously performing part of the runs.If the possible triplets are big,please work on a big memory machine.

Examples

##Different types of candidate modulators to predict
##Here we take four candidate modulators for example
##Two for modulator;one for initialnot;one for filterdnot(see value section in details)
tri.app(ms=datatests[["m_app"]],ET=datatests[["ET"]],M.exp=datatests[["M_exp"]],
        E.exp=datatests[["E_exp"]],T.exp=datatests[["T_exp"]])