calculateLpeAdj: Evaluates local pooled error significance test with user...

In LPEadj: A correction of the local pooled error (LPE) method to replace the asymptotic variance adjustment with an unbiased adjustment based on sample size.

Description

The local pooled error test attempts to reduce dependence on the within-gene estimates in tests for differential expression, by pooling error estimates within regions of similar intensity. Note that with the large number of genes there will be genes with low within-gene error estimates by chance, so that some signal-to-noise ratios will be large regardless of mean expression intensities and fold-change. The local pooled error attempts to avert this by combining within-gene error estimates with those of genes with similar expression intensity.

Usage

  calculateLpeAdj(x, y, basevar.x,basevar.y, df=10, array.type="olig", 
      probe.set.name="OLIG.probe.name", trim.percent=5, adjust1=1.57, adjust2=1.57)

Arguments

x	Replicated data from first experimental condition (as matrix or data-frame)

.

y	Replicated data from second experimental condition (as matrix or data-frame)

.

basevar.x	Baseline distribution of first condition obtained from function baseOlig.error
basevar.y	Baseline distribution of second condition obtained from function baseOlig.error
df	Degrees of freedom used in fitting smooth.spline to estimates of var.M for bins in A
array.type	Currently supports oligo arrays
probe.set.name	Gene IDs. By default if they are not provided then 1,2,3,... is assigned as GeneID
trim.percent	Percent of (A, var.M) estimates to trim from low end of A
adjust1	adjustment factor of variance for first group
adjust2	adjustment factor of variance for second group

Details

The LPE test statistic numerator is the difference in medians between the two experimental conditions. The test statistic denominator is the combined pooled standard error for the two experimental conditions obtained by looking up the var.M from each baseOlig.error variance function. The conversion to p-values is based on the Gaussian distribution for difference if order statistics (medians). The user may select both the smoother degrees of freedom (smaller is smoother) and the trim percent to obtain a variance function to suit particular issues i.e. variability of genes with low expression intensity. The default values for the adjustment of the variances of the two groups is the asymptotically correct value of pi/2. This value is biased at small sample values and unbiased adjustment parameters based on sample size can be used instead. See documentation of lpeAdj for details.

Value

Data frame including x, median of x, y, median of y, median difference of (x,y), pooled standard deviation of difference, LPE p-value, outlier flag, probability of an outlier within x or y.

Author(s)

Carl Murie [e-mail], Nitin Jain [e-mail]

References

J.K. Lee and M.O.Connell(2003). An S-Plus library for the analysis of differential expression. In The Analysis of Gene Expression Data: Methods and Software. Edited by G. Parmigiani, ES Garrett, RA Irizarry ad SL Zegar. Springer, NewYork.

Jain et. al. (2003) Local pooled error test for identifying differentially expressed genes with a small number of replicated microarrays, Bioinformatics, 1945-1951.

Jain et. al. (2005) Rank-invariant resampling based estimation of false discovery rate for analysis of small sample microarray data, BMC Bioinformatics, Vol 6, 187.

See Also

lpeAdj

Examples

   
 # Loading the data from the LPE library
 data(Ley)

 ADJ.VALUES <- c(1, 1, 1.34585905516761 ,1.19363228146169 ,1.436849413109
                  ,1.289652132873 ,1.47658053092781 ,1.34382984852146
                  ,1.49972130857404, 1.3835405678718)
            
 dim(Ley)
 # Gives 12488*7 
 # First column is ID.



 # Subsetting the data
 subset.Ley <- Ley[1:1000,]
  
 subset.Ley[,2:7] <- preprocess(subset.Ley[,2:7],data.type="MAS5")
  
 # Finding the baseline distribution of condition 1 and 2.
 var.1 <- adjBaseOlig.error(subset.Ley[,2:4], q=0.01, setMax1=FALSE)
 var.2 <- adjBaseOlig.error(subset.Ley[,5:7], q=0.01, setMax1=FALSE)
 
 # Applying LPE
 lpe.result <- calculateLpeAdj(subset.Ley[,2:4],subset.Ley[,5:7], var.1, var.2,
		probe.set.name=subset.Ley[,1], adjust1=ADJ.VALUES[3],
                adjust2=ADJ.VALUES[3])
  

Example output

Loading required package: LPE
[1] 12488     7

LPEadj documentation built on Nov. 8, 2020, 8:29 p.m.