GeneReg-package: Construct time delay gene regulatory network

Description Details Author(s) References

Description

GeneReg is an R package for inferring time delay gene regulatory network using time course gene expression data.

Details

The goal of time delay linear model is to fit a linear regression model using a set of putative regulators to estimate the transcription pattern of a specific target gene.

Eg = a1*Etf1(-t1) + a2*Etf2(-t2) + ... + ai*Etfi(-ti) + ... + an*Etfn(-tn)

Where Eg is the relative expression level of target gene, Etfi is the relative expression level of TFi, ti is time delay of TFi, ai is the regression coefficient of TFi.

The regression models were optimized by the AIC forward selection procedure.

In the following example, we will demonstrate how to use this package to analysis a cell cycle datasets which is publicly available at GEO http://www.ncbi.nlm.nih.gov/geo under accession number GSE8799. The data set included the gene expression profiles of wild-type cells and cyclin-mutant cells with 15 time points during two cell cycles. 1271 periodic genes acquired from wild-type cells were analyzed to investigate how yeast regulators regulate target gene expression. These periodic genes formed the list of target genes. Then, a candidate pool of potential regulators of target genes was constructed by intersecting the periodic genes and the transcription factors in YEASTRACT http://www.yeastract.com/.

First, B spline interpolation was applied to estimate 100 time points according to the original 15 time points.

> data(wt.expr.data)

> wt.bspline.data<- ts.bspline(wt.expr.data,

+ ts.point=as.numeric(colnames(wt.expr.data)), data.predict=100)

Then, the time delay linear model was carried out based on the interpolated expression data. single.adj.r.squared and multiple .adj.r.squared were the cutoffs of adjusted R seqaured of single regulator regression and multiple regulator regression, respectively. tfs.list were the candidate pool of regulators. min.coef means the regression coefficient should be larger than 1/4 and smaller than 4. maxdelay means the time points delay should not exceed 1/4 of all time points.

> data(tf.list)

> dir.create('wt.model')

> setwd('wt.model')

> wt.models<-timedelay.lm.batch(bspline.data=wt.bspline.data,

+ expr.data=wt.expr.data, regulator.list=tf.list,

+ target.list=rownames(wt.bspline.data),

+ single.adj.r.squared=0.8, multiple.adj.r.squared=0.9,

+ maxdelay=ncol(wt.bspline.data)*0.1, min.coef=0.25, max.coef=4,

+ output=T, topdf=T, xlab='Time point (lifeline)',

+ ylab='Relative expression level (in log ratio)')

> setwd('..')

Finally, the gene regulatory network can be plotted according to the series of time delay linear models.

> pdf('wt.network.pdf',width=70, height=70)

> plot.GeneReg(wt.models,vertex.size=2,layout=layout.fruchterman.reingold)

> dev.off()

The analysis of cyclin-mutant cells is similar.

Author(s)

Tao Huang <[email protected]>

References

Tao Huang, Lei Liu, Ziliang Qian, Kang Tu, Yixue Li, Lu Xie: Using GeneReg to construct time delay gene regulatory networks. BMC Res Notes 2010, 3(1):142. http://www.biomedcentral.com/1756-0500/3/142


GeneReg documentation built on May 30, 2017, 6:07 a.m.