plotOnGene: Plot probe level data on a gene

Description

Function that will investigate the probe level intensity of probes as a function of their location in a gene.

Usage

plotOnGene(object, gene, probeData=NULL, label=NULL, genename=NULL, summaryType="median",
    interval=NULL, ylim=NULL, testType=NULL, forcePValue=FALSE, verbose=TRUE, cutoff=0.2, directions="all", ylab="expression")

Arguments

object

A ProbeLevelSet object or a regular ExpressionSet object (in which case a probeData argument is required). See getLocalProbeIntensities and related functions on how to create a ProbeLevelSet.

gene

A number of gene sequences as DNAstring, DNAStringSets, or character-vectors with sequence.

probeData

Optional if a ProbeLevelSet is submitted as object argument. Otherwise, it must be a data frame with rownames corresponding to the featureNames of the ExpressionSet and a column named "sequence" with the probe sequences as character strings

label

An optional character string specifying a column name in the pData of the object. If this argument is given, the gene plot will be colour coded based on the different groups (factors) in the pData entry. If a summaryType other than 'dots' is selected the summarisation is done stratified by the different groups in the pData. It can be a numeric or integer entry, but it will be coerced to factors.

genename

Optional character string specifying a gene name to include in the plot. If not included and a FASTA sequence is given, it will default to the name in the FASTA sequence. Otherwise, it will default to 'Unknown genename'.

summaryType

Character string specifying one of the following summary methods: 'median', 'mean', 'quartiles' or 'dots' (i.e. no summary). Specifies how all the sample values or all the samples values in a group if 'label' is given, should be summarised. Defaults to 'median'.

interval

Optional vector of two integers of bp positions. If given, the plot will only include the sequence from gene in the given interval. The x-axis annotation is preserved from original, so this is useful for zooming on specific regions.

ylim

Optional two integers. If given, this value will be the minimal and maximal value on the y-axis. This is useful if a few outlier probes have very high intensity values, as the default is to set the ylim from the maximal intensity value.

testType

Optional character string, defining a statistic procedure to identify especially interesting probes. Can be either 'linear model', 'students' or 'wilcoxons'. If given, a label must also be specified. In this case the plotStatistics function will be called and probes that are significantly changed between the groups in label at the P-value set in cutoff (see cutoff argument) will be circled.

forcePValue

Logical. Is used if the testType argument is used. If TRUE all significantly changed probes have P-value given on the plot. If FALSE, only plots with less than 10 significant probes write P-values. Plots can become very cluttered with data if set to TRUE

verbose

TRUE or FALSE

cutoff

Integer specifying at what p-value probes should be circled when using the 'testType' variable. Defaults to 0.2. For cutoffs higher than 0.05, all probes with P >0.05 will be circled in grey instead of black.

directions

A character vector of the matching-directions that should be scanned (which combinations of complementary and reverse). Defaults to "all" which is shorthand for all possible directions, but can take anything from: c("matchForwardSense", "matchForwardAntisense", "matchReverseSense", "matchReverseAntisense")

ylab

Label of Y-axis as in default plots

Value

No value, but plots the local expression levels relations of each probe found in the submitted gene sequence as a function of its location along this sequence. Various statistics and summarizations on pdata can be employed, as specified in details.

Details

At the very least, this function takes a ProbeLevelSet or an ExpressionSet + probeData and the sequence of a gene. It then compares the probe sequences given in the ProbeLevelSet or the probeData variable with the sequence of the gene given. Any probes with sequences found in the gene will be plotted, with their intensity level on the y-axis and their location in the gene on the x-axis. If no further arguments are given this gives a view of relative expression levels along the length of the gene, and can be used to investigate which exons are actively transcribed in the sample and which are not. An important argument that can be used for further investigation is the 'label' argument which specifies a column in the pData of the ExpressionSet. In this case the plots will be stratified by the factors specified in this column (so giving labels with numerical or Date class data will not work). This can be very useful when investigating how different sample conditions affect various regions of a gene. A transcript isoform that is relatively upregulated in a diseased state will for example not be discovered if a probeset or metaprobeset covering the entire gene is used to summarize the data, since the average expression intensity for the gene will remain constant. Using the plotOnGene function, however, and specifying a case / control label will reveal tendencies for probes at certain exon locations to have relation to this label. The testType argument further supports this functionality by providing statistical testing and highlighting of probes that correlate significantly to the given label. In the case / control example, a student's t-test would highlight all probes that matched with the exons of the gene that was only found in the disease-specific transcript isoform. When interpreting the data it is suggested that specific attention is paid to the pattern of probes in the same exon. A single probe with a P-value < 0.05 might be a false positive caused by chance or by cross hybridization of the probe sequence to something else. A range of probes in the same exon that all show P-values below or close to 0.05, however, is much more likely to be an actual case of a transcript isoform having this particular exon or exons being regulated between the groups in the label. Exon structure can be easily plotted on the graph using the exonStructure function. A special case is the search for SNPs which have effect on expression levels or variable splicing. The testType argument 'linear model' is designed for this. The linear model calls the internal function doProbeLinear which assign each of the levels in the 'label' column of the pdata a value between 1 and the number of levels, in the order in which they are sorted. For genotypes given as "AA", "AB", "BB" character strings this will give "AA" = 1, "AB" = 2 and "BB" = 3. The doProbeLinear then calculates a linear model between the intensity values and these numbers, and returns the P-value. In this case, low P-values can be interpreted as a case where the heterozygote samples have intermediary expression levels between the two homozygotes. This is the case that can be expected to be seen if the nucleotide type of SNP does in fact have any effect on the mRNA concentration levels in the sample.

Examples

Run this code

data(exampleProbeLevelSet)
	
	plotOnGene(exampleProbeLevelSet, mrna, summaryType="dots", interval=c(500,1000))
	
	plotOnGene(exampleProbeLevelSet, mrna, label="genotype3", testType="linear model")
	exonStructure(mrna, genomic)