Bne-a-ms5.retain.aarnet.edu.au

Title Differential Binding Analysis of ChIP-Seq peak data Rory Stark<rory.stark@cancer.org.uk>, Gordon Brown <gordon.brown@cancer.org.uk> Maintainer Rory Stark<rory.stark@cancer.org.uk> Description Compute differentially bound sites from multiple ChIP-seq experiments using affin- ity (quantitative) data. Also enables occupancy (overlap) analysis and plotting functions.
biocViews Bioinformatics, HighThroughputSequencing, ChIPseq Imports RColorBrewer, amap, edgeR (>= 2.3.58), gplots, limma, DE- Seq,grDevices, stats, utils, IRanges, zlibbioc Collate core.R parallel.R counts.R contrast.R analyze.R io.R helper.R utils.R DBA.R DiffBind-package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
dba . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
DBA object methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
DBA tamoxifen resistance dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
dba.analyze . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
dba.contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
dba.count . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.peakset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.plotBox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.plotHeatmap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.plotMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.plotPCA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.plotVenn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.save . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . dba.show . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . DiffBind – DBA global constant variables . . . . . . . . . . . . . . . . . . . . . . . . . Differential Binding Analysis of ChIP-seq peaksets Differential binding analysis of ChIP-seq peaksets Computes differentially bound sites from multiple ChIP-seq experiments using affinity (quantita-tive) data. Also enables occupancy (overlap) analysis and plotting functions.
Add a peakset to, or retrieve a peakset from, a dba object Compute binding site overlaps and/or correlations Rory Stark <rory.stark@cancer.org.uk> and Gordon Brown <gordon.brown@cancer.org.uk> Constructs a new DBA object from a sample sheet, or based on an existing DBA object sampleSheet="dba_samples.csv",config=data.frame(RunParallel=TRUE, reportInit="DBA"),caller=’raw’, skipLines=0, bAddCallerConsensus=FALSE,bRemoveM=TRUE, bRemoveRandom=TRUE,bCorPlot=FALSE, attributes) existing DBA object – if present, will return a fully-constructed DBA objectbased on the passed one, using criteria specified in the mask and/or minOverlapparameters. If missing, will create a new DBA object based on the sampleSheet.
logical or numerical vetcor indicating which peaksets to include in the resultingmodel if basing DBA object on an existing one. See dba.mask.
only include peaks in at least this many peaksets in the main binding matrix ifbasing DBA object on an existing one.
data frame containing sample sheet, or file name of sample sheet to load (ignoredif DBA is specified). Columns names in sample sheet should include: • SampleID: Identifier string for sample • Tissue: Identifier string for tissue type • Condition: Identifier string for condition • Replicate: Replicate number of sample • bamReads: file path for bam file containing aligned reads for ChIP sample • bamControl: file path for bam file containing aligned reads for control sam- • ControlID: Identifier string for control sample (optional) • Peaks: path for file containing peaks for sample. format determined by • PeakCaller: Identifier string for peak caller used. If Peaks is not a bed file, this will determine how the Peaks file is parsed. If missing, will use defaultpeak caller specified in caller parameter. Possible values: – “raw”: text file file; peak score is in fourth column – “bed”: .bed file; peak score is in fifth column – “peakset”: peakset written out using pv.writepeakset data frame containing sample sheet, or file name of config file to load whenconstructing a new DBA object from a sample sheet. NULL indicates no configfile. Relevant fields include: • RunParallel: logical indicating if counting and analysis operations should be run in parallel using multicore by default.
• DataType: default class for peaks and reports (DBA_DATA_GRANGES, DBA_DATA_RANGEDDATA, or DBA_DATA_FRAME).
• AnalysisMethod: either DBA_EDGER or DBA_DESEQ.
if a sampleSheet is specified, the default peak file format that will be used if thePeakCaller column is absent.
if a sampleSheet is specified, the number of lines (ie header lines) at the begin-ning of each peak file to skip.
add a consensus peakset for each sample with more than one peakset (i.e. dif-ferent peak callers) when constructing a new DBA object from a sample sheet.
logical indicating whether to remove peaks on chrM (mitochondria) when con-structing a new DBA object from a sample sheet.
logical indicating whether to remove peaks on chrN_random when constructinga new DBA object from a sample sheet.
logical indicating that a correlation heatmap should be plotted before returning vector of attributes to use subsequently as defaults when generating labels inplotting functions: MODE: Construct a new DBA object from a samplesheet: dba(sampleSheet, config, bAddCallerConsensus, bRemoveM, bRemoveRandom, attributes) MODE: Construct a DBA object based on an existing one: # Create DBA object from a samplesheetsetwd(system.file("extra", package="DiffBind"))tamoxifen = dba(sampleSheet="tamoxifen.csv")tamoxifen #Create a DBA object with a subset of samplesdata(tamoxifen_peaks)Responsive = dba(tamoxifen,tamoxifen$masks$Responsive)Responsive ## S3 method for class ’DBA’print(x, .)## S3 method for class ’DBA’summary(object, .)## S3 method for class ’DBA’plot(x, .) data(tamoxifen_peaks)tamoxifendata(tamoxifen_counts)tamoxifen Tamoxifen resistance dataset used for DBA examples Tamoxifen resistance dataset used for DBA examples load tamoxifen resistance dataset DBA object with peak (occupancy) data load tamoxifen resistance dataset DBA object with count (affinity) data load tamoxifen resistance dataset DBA object with count (affinity) data andedgeR-based differential binding analysis results The tamoxifen resistance dataset is used for the DBA vignette and man page examples.
data(tamoxifen_peaks)tamoxifendata(tamoxifen_counts)plot(tamoxifen)data(tamoxifen_analysis)dba.plotMA(tamoxifen) Perform differential binding affinity analysis Performs differential binding affinity analysis dba.analyze(DBA, method=DBA$config$AnalysisMethod, bSubControl=TRUE, bFullLibrarySize=FALSE, bTagwise=TRUE,bCorPlot=TRUE, bReduceObjects=T, bParallel=DBA$config$RunParallel) DBA object. If no contrasts are specified (DBA$contrast is NULL), defaultcontrasts will be added via a call to dba.contrast(DBA).
method, or vector of methods, by which to analyze differential binding affinity.
Supported methods: logical indicating whether Control read counts are subtracted for each site ineach sample before performing analysis.
logical indicating if the full library size (total number of reads in BAM/SAM/BEDfile) for each sample is used for scaling normalization. If FALSE, the total num-ber of reads present in the peaks for each sample is used (generally preferable).
logical indicating if dispersion should be calculated on a tagwise (or per-condition)basis. If there are only a very few members of each group in a contrast (e.g. noreplicates), this should be set to FALSE.
logical indicating whether to plot a correlation heatmap for the analyzed data(first contrast only). If no sites are significantly differentially bound using thedefault threholds, no heatmap will be plotted.
bReduceObjects logical indicating whether strip the analysis objects of unnecessary fields to save memory. If it is desired to used the DBA$contrasts[[n]]$edgeR and/orDBA$contrasts[[n]]$DESeq objects directly in the edgeR and/or DESeq pack-ages, this should be set to FALSE.
logical indicating that the analyses is to be done in parallel using multicore(one process for each contrast for each method, plus an additional process permethod).
See the DBA User Guide for more details on how the edgeR and DESeq analyses are carried out.
DBA object with results of analysis added to DBA$contrasts.
If the "edgeR" method is specified, and there is a blocking factor for the contrast(s) specified usinga previous call to dba.contrast, a multi-factor analysis will automatically be carried out in additionto a single factor analysis.
tamoxifen = dba.analyze(tamoxifen)tamoxifen tamoxifen = dba.analyze(tamoxifen,method=c(DBA_EDGER,DBA_DESEQ))tamoxifen Set up contrasts for differential binding affinity analysis Sets up contrasts for differential binding affinity analysis dba.contrast(DBA, group1, group2=!group1, name1="group1", name2="group2", minMembers=3, block ,categories = c(DBA_TISSUE,DBA_FACTOR,DBA_CONDITION,DBA_TREATMENT)) mask of samples in first group (when adding a specific contrast). See dba.mask mask of samples in second group (when adding a specific contrast). See dba.mask label for samples in first group (when adding a specific contrast).
label for samples in second group (when adding a specific contrast).
when automatically generating contrasts, minimum number of unique samplesin a group. Must be at least 2, as replicates are strongly advised. If you wish todo an analysis with no replicates, you can set the group1 and group2 parametersexplicitly.
when automatically generating contrasts, attribute or vector of attributes to basecontrasts on: blocking attribute for multi-factor analysis. This may be specified as either avalue, a vector, or a list.
If block is a value, the specified metadata field is used to derive the blockingfactor. One of: If block is a vector, it can either be a mask (logical vector) or a vector of peaksetnumbers. In this case, the peaksets indicated in the blocking vector are all giventhe same value (true), while any peaksets not included in the vector take thealternative value (false).
If block is a list, it should be a list of vectors (either logical masks or vectorsof peakset numbers), with each indicating a set of peaksets that should sharethe same value. Each peasket should appear at most once, and any peaksets notspecified will be given an default value (other).
dba.contrast(DBA, minMembers, categories) dba.contrast(DBA, group1, group2, name1, name2, block) DBA object with contrast(s) set as DBA$contrasts. Contrast list can be retrieved using dba.show(DBA,bContrasts=T).
Contrasts will only be set up for peaksets where DBA_CALLER == "counts".
Contrasts can be cleared by DBA$contrasts=NULL.
data(tamoxifen_counts)tamoxifen = dba.contrast(tamoxifen, categories=DBA_CONDITION)tamoxifen # Another way to do the same thingtamoxifen$contrasts=NULLtamoxifen = dba.contrast(tamoxifen, tamoxifen$masks$Responsive, tamoxifen$masks$Resistant, "Responsive", "Resistant") # Add add default contraststamoxifen$contrasts=NULLtamoxifen = dba.contrast(tamoxifen)tamoxifen # Specify a blocking factortamoxifen$contrasts=NULLtamoxifen = dba.contrast(tamoxifen, categories=DBA_CONDITION, block=DBA_TISSUE)tamoxifen tamoxifen$contrasts=NULLtamoxifen = dba.contrast(tamoxifen, categories=DBA_CONDITION, block=list(c(3,4,5,8,9),c(1,2,10,11)))tamoxifen tamoxifen$contrasts=NULLtamoxifen = dba.contrast(tamoxifen, categories=DBA_CONDITION, block=tamoxifen$masks$MCF7)tamoxifen = dba.analyze(tamoxifen)tamoxifen Counts reads in binding site intervals. Files must be one of bam, bed and gzip-compressed bed. Filesuffixes must be ".bam", ".bed", or ".bed.gz" respectively.
dba.count(DBA, peaks, minOverlap=2, score=DBA_SCORE_TMM_MINUS_EFFECTIVE, bLog=FALSE, insertLength, maxFilter, bRemoveDuplicates=FALSE, bCalledMasks=TRUE,bCorPlot=TRUE, bParallel=DBA$config$RunParallel) GRanges, RangedData, dataframe, or matrix containing intervals to use. Ifmissing, generates a consensus peakset using minOverlap parameter. If NULL,changes the score used in the global binding matrix to the score type specifiedin the score parameter.
only include peaks in at least this many peaksets when generating consensuspeakset (i.e. when peaks parameter is missing).
which score to use in the binding affinity matrix. Note that all raw read countsare maintained for use by dba.analyze, regardless of how this is set. One of: raw read count for interval using only reads from ChIP raw read count for interval from ChIP divided by read count for interval from control raw read count for interval from ChIP minus read count for interval from control RPKM for interval using only reads from ChIP RPKM for interval from ChIP divided by RPKM for interval from control TMM normalized (using edgeR), using ChIP read counts and Full Library size TMM normalized (using edgeR), using ChIP read counts and Effective Library size TMM normalized (using edgeR), using ChIP read counts minus Control read counts and Full Library size TMM normalized (using edgeR), using ChIP read counts minus Control read counts and Effective Library size logical indicating whether log2 of score should be used (only applies to DBA_SCORE_RPKM_FOLDand DBA_SCORE_READS_FOLD).
if present, this value will be used as the length of the reads. Each read will beextended from its endpoint along the appropriate strand by this many bases. Ifmissing, the read size indicated in the BAM/SAM/BED file will be used.
value to use for filtering intervals with low read counts. Only intervals whereat least one sample has at least maxFilter reads will be included. If missing,includes all intervals. If peaks is NULL, will remove sites from existing DBAobject without recounting.
logical indicating if duplicate reads (ones that map to exactly the same genomicposition) should be removed. If TRUE, any location where multiple reads mapwill be counted as a single read.
logical indicating whether to compute site masks for each peakset indicatingwhich sites were originally identified as peaks(used by dba.report).
logical indicating whether to plot a correlation heatmap for the counted data if TRUE, use multicore to get counts for each read file in parallel DBA object with binding affinity matrix based on read count scores.
# These won’t run unless you have the reads available in a BAM, SAM, or BED filedata(tamoxifen_peaks)## Not run: tamoxifen = dba.count(tamoxifen) # Count using a peakset made up of only peaks in all responsive MCF7 replicatesdata(tamoxifen_peaks)mcf7Common = dba.overlap(tamoxifen,tamoxifen$masks$MCF7&tamoxifen$masks$Responsive) ## Not run: tamoxifen = dba.count(tamoxifen,peaks=mcf7Common$inAll)tamoxifen # Change binding affinity scoresdata(tamoxifen_counts)tamoxifen = dba.count(tamoxifen,peaks=NULL,score=DBA_SCORE_READS)head(tamoxifen$vectors)tamoxifen = dba.count(tamoxifen,peaks=NULL,score=DBA_SCORE_RPKM_FOLD)head(tamoxifen$vectors)tamoxifen = dba.count(tamoxifen,peaks=NULL,score=DBA_SCORE_TMM_MINUS_FULL)head(tamoxifen$vectors) dba.load(file=’DBA’, dir=’.’, pre=’dba_’, ext=’RData’) data(tamoxifen_peaks)dba.save(tamoxifen,’tamoxifenPeaks’)tamoxifen = dba.load(’tamoxifenPeaks’) Derive a mask to define a subset of peaksets or sites for a DBA object Derives a mask to define a subset of peaksets or sites for a DBA object.
dba.mask(DBA, attribute, value, combine=’or’, mask, merge=’or’, bApply=FALSE, when deriving a peakset mask, attribute to base mask on: when deriving a peakset/sample mask, attribute value (or vector of attribute val-ues) to match.
when deriving a peakset/sample mask, if value is a vector, OR when deriving asite mask, and peaksets is a vector, this is method for combining result of eachvalue: when deriving a peakset/sample mask, this specifies an existing mask to mergewith; if missing, create new mask when deriving a peakset/sample mask, and an existing mask is supplied, thisspeficies the method for combining new mask with supplied mask: • “nand” note: if mask is missing, “nand” results in negative of mask when deriving a peakset/sample mask, a logical indicating that a new DBA ob-ject with the mask applied will be returned.
when deriving a peak/site mask, this specifies a peakset number, or a vector ofpeakset numbers. The resulting mask will indicate which of the overall siteswere called as peaks in this peakset or set of peaksets. If a vector, the masks foreach of the peaksets will be combined using the method specified in the combineparameter.
when deriving a peak/site mask, scores greater than this value will be consideredas indicating that the site corresponds to a called peakset.
MODE: Derive a a mask of peaksets/samples: dba.mask(DBA, attribute, value, combine, mask, merge, bApply) dba.mask(DBA, combine, mask, merge,bApply, peakset, minValue) either a logical mask, or new DBA object if bApply is TRUE.
dba automatically generates masks for each unique value of DBA_TISSUE, DBA_FACTOR, DBA_CONDITION,DBA_TREATMENT, DBA_CALLER, and DBA_REPLICATE. These are accessible using masksfield of the DBA object (DBA$masks), and can be viewed using names(DBA$masks).
# Pre-made masksnames(tamoxifen$masks)dba.show(tamoxifen,tamoxifen$masks$MCF7) # New masksmcf7Mask = dba.mask(tamoxifen,DBA_TISSUE, "MCF7")mcf7DerivedMask = dba.mask(tamoxifen,DBA_TISSUE,"TAMR",mask=mcf7Mask)mcf7Derived = dba(tamoxifen,mcf7DerivedMask)mcf7Derived Compute binding site overlaps (occupancy analysis) Computes binding overlaps and co-occupancy statistics dba.overlap(DBA, mask, mode=DBA_OLAP_PEAKS, minVal=0, contrast, method=DBA$config$AnalysisMethod, th=.1, bUsePval=FALSE,report, byAttribute, bCorOnly=TRUE, CorMethod="pearson",DataType=DBA$config$DataType) mask or vector of peakset numbers indicating a subset of peaksets to use (seedba.mask). When generating overlapping/unique peaksets, either two or threepeaksets must be specified. If the mode type is DBA_OLAP_ALL, and a con-trast is specified, a value of TRUE (mask=TRUE) indicates that all samplesshould be included (otherwise only those present in one of the contrast groupswill be included).
indicates which results should be returned (see MODES below). One of: minimum score value to be considered a "called" peak.
contrast number to use. Only specified if contrast data is to be used whenmode=DBA_OLAP_ALL. See dba.show(DBA, bContrast=T) to get contrast num-bers.
if contrast is specified and mode=DBA_OLAP_ALL, use data from method usedfor analysis: if contrast is specified and mode=DBA_OLAP_ALL, significance threshold; allsites with FDR (or p-values, see bUsePval) less than or equal to this value willbe included. A value of 1 will include all binding sites, but only the samplesincluded in the contrast.
if contrast is specified and mode=DBA_OLAP_ALL, logical indicating whetherto use FDR (FALSE) or p-value (TRUE) for thresholding.
if contrast is specified and mode=DBA_OLAP_ALL, a report (obtained fromdba.report) specifying the data to be used. If counts are included in the report(and a contrast is specified), the count data from the report will be used to com-pute correlations, rather than the scores in the global binding affinity matrix. Ifreport is present, the method, th, and bUsePval parameters are ignored.
when computing co-occupancy statistics (DBA_OLAP_ALL), limit compar-isons to peaksets with the same value for a specific attribute, one of: when computing co-occupancy statistics (DBA_OLAP_ALL), logical indicat-ing that only correlations, and not overlaps, should be computed. This is muchfaster if only correlations are desired (e.g. to plot the correlations using dba.plotHeatmap).
when computing co-occupancy statistics (DBA_OLAP_ALL), method to usewhen computing correlations.
if mode==DBA_OLAP_PEAKS, the class of object that peaksets should be re-turned as: Can be set as default behavior by setting DBA$config$DataType.
MODE: Generate overlapping/unique peaksets: dba.overlap(DBA, mask, mode=DBA_OLAP_PEAKS, minVal) MODE: Compute correlation and co-occupancy statistics (e.g. for dba.plotHeatmap): dba.overlap(DBA, mask, mode=DBA_OLAP_ALL, byAttribute, minVal, attributes, bCorOnly, CorMethod) MODE: Compute correlation and co-occupancy statistics using significantly differentially boundsites (e.g. for dba.plotHeatmap): dba.overlap(DBA, mask, mode=DBA_OLAP_ALL, byAttribute, minVal, contrast, method, th=,bUsePval, attributes, bCorOnly, CorMethod) Note that the scores from the global binding affinity matrix will be used for correlations unless areport containing count data is specified.
MODE: Compute overlap rates at different stringency thresholds: dba.overlap(DBA, mask, mode=DBA_OLAP_RATE, minVal) Value depends on the mode specified in the mode parameter.
If mode = DBA_OLAP_PEAKS, Value is an overlap record: a list of three peaksets for an A-Boverlap, and seven peaksets for a A-B-C overlap: peaks in both peaksets B and C but not peakset A peaks in both peaksets A and C but not peakset B peaks in both peaksets A and B but not peakset C If mode = DBA_OLAP_ALL, Value is a correlation record: a matrix with a row for each pair ofpeaksets and the following columns: peakset number of first peakset in overlap peakset number of second peakset in overlap number of peaks in both peakset A and B (merged) percentage overlap (number of overlapping sites divided by number of peaksunique to smaller peakset If mode = DBA_OLAP_RATE, Value is a vector whose length is the number of peaksets, contain-ing the number of overlapping peaks at the corresponding minOverlaps threshold (i.e., Value[1] isthe total number of unique sites, Value[2] is the number of unique sites appearing in at least twopeaksets, Value[3] the number of sites overlapping in at least three peaksets, etc.).
data(tamoxifen_peaks)# default mode: DBA_OLAP_PEAKS -- get overlapping/non overlapping peaksetsmcf7 = dba.overlap(tamoxifen,tamoxifen$masks$MCF7&tamoxifen$masks$Responsive)names(mcf7)mcf7$inAll mcf7 = dba(tamoxifen,tamoxifen$masks$MCF7)mcf7.corRec = dba.overlap(mcf7,mode=DBA_OLAP_ALL,bCorOnly=FALSE)mcf7.corRec # mode: DBA_OLAP_RATE -- get overlap rate vectordata(tamoxifen_peaks)rate = dba.overlap(tamoxifen, mode=DBA_OLAP_RATE)rateplot(rate,type=’b’,xlab="# peaksets",ylab="# common peaks", main="Tamoxifen dataset overlap rate") Add a peakset to, or retrieve a peakset from, a DBA object Adds a peakset to, or retrieves a peakset from, a DBA object dba.peakset(DBA=NULL, peaks, sampID, tissue, factor, condition, treatment, replicate, control, peak.caller, reads=0, consensus=FALSE,bamReads, bamControl,normCol=4, bRemoveM=TRUE, bRemoveRandom=TRUE,minOverlap=2, bMerge=TRUE,bRetrieve=FALSE, writeFile, numCols=4,DataType=DBA$config$DataType) DBA object. Required unless creating a new DBA object by adding an initialpeakset.
When adding a specified peakset: set of peaks, either a GRanges or RangedDataobject, or a peak dataframe or matrix (chr,start,end,score), or a filename wherethe peaks are stored.
When adding a consensus peakset: a sample mask or vector of peakset numbers.
If missing or NULL, a consensus is derived from all peaksets present in themodel. See dba.mask, or dba.show to get peakset numbers.
When adding all the peaks from one DBA object to another: a DBA object. Inthis case, the only other parameter to have an effect is minOverlap.
When retrieving and/or writing a peakset: either a GRanges or RangedData ob-ject, or a peak dataframe or matrix (chr,start,end,score), or a peakset number; ifNULL, retrieves/writes the full binding matrix.
ID string for the peakset being added; if missing, one is assigned (a serial num-ber for a new peakset, or a concatenation of IDs for a consensus peakset).
tissue name for the peakset being added; if missing, one is assigned for a con-sensus peakset (a concatenation of tissues).
factor name for the peakset being added; if missing, one is assigned for a con-sensus peakset (a concatenation of factors).
condition name for the peakset being added; if missing, one is assigned for aconsensus peakset (a concatenation of conditions).
treatment name for the peakset being added; if missing, one is assigned for aconsensus peakset (a concatenation of treatment).
replicate number for the peakset being added; if missing, one is assigned for aconsensus peakset (a concatenation of replicate numbers).
control name for the peakset being added; if missing, one is assigned for a con-sensus peakset (a concatenation of control names).
peak caller name string. If peaks is specified as a file, this will control how it isinterpreted. Supported values: • “raw”: text file file; peak score is in fourth column • “bed”: .bed file; peak score is in fifth column • “peakset”: peakset written out using pv.writepeakset if missing, a name is assigned for a consensus peakset (a concatenation of peakcaller names).
total number of ChIPed library reads for the peakset being added.
TRUE if peakset being added is made from overlap of other peaksets (set auto-matically when adding a consensus peakset).
file path of the BAM/SAM/BED file containing the aligned reads for the peaksetbeing added.
file path of the BAM/SAM/BED file containing the aligned reads for the controlused for the peakset being added.
peak column to normalize to 0.1 scale when adding a peakset; 0 indicates nonormalization logical indicating whether to remove peaks on chrM when adding a peakset logical indicating whether to remove peaks on chrN_random when adding apeakset the minimum number of peaksets a peak must be in to be included when addinga consensus peakset. When retrieving, if the peaks parameter is a vector (logicalmask or vector of peakset numbers), a binding matrix will be retrieved includingall peaks in at least this many peaksets.
logical indicating whether global binding matrix should be compiled after addingthe peakset. When adding several peaksets via successive calls to dba.peakset,it may be more efficient to set this parameter to FALSE and call dba(DBA) afterall the peaksets have been added.
logical indicating that a peakset is being retrieved and/or written, not added.
number of columns to include when writing out peakset. First four columns arechr, start, end, score; the remainder are maintained from the original peakset.
Ignored when writing out complete binding matrix.
The class of object for returned peaksets: Can be set as default behavior by setting DBA$config$DataType.
dba.peakset(DBA=NULL, peaks, sampID, tissue, factor, condition, replicate, control, peak.caller,reads, consensus, bamReads, bamControl, normCol, bRemoveM, bRemoveRandom) MODE: Add a consensus peakset (derived from overlapping peaks in peaksets already present): dba.peakset(DBA, peaks, bRetrieve=T, writeFile, numCols) DBA object when adding a peakset. Peakset matrix or RangedData object when retrieving and/orwriting a peakset.
# create a new DBA object by adding three peaksetsmcf7 = dba.peakset(NULL, peaks=system.file("extra/peaks/MCF7_ER_1.bed.gz", package="DiffBind"),sampID="MCF7.1",tissue="MCF7",factor="ER",condition="Responsive",replicate=1) peaks=system.file("extra/peaks/MCF7_ER_2.bed.gz", package="DiffBind"),sampID="MCF7.2",tissue="MCF7",factor="ER",condition="Responsive",replicate=2) peaks=system.file("extra/peaks/MCF7_ER_3.bed.gz", package="DiffBind"),sampID="MCF7.3",tissue="MCF7",factor="ER",condition="Responsive",replicate=3) #retrieve peaks that are in all three peaksetsmcf7.consensus = dba.peakset(mcf7, 1:3, minOverlap=3, bRetrieve=TRUE)mcf7.consensus #add a consensus peakset -- peaks in all three replicatesmcf7 = dba.peakset(mcf7, 1:3, minOverlap=3,sampID="MCF7_3of3")mcf7 #retrieve the consensus peakset as RangedData objectmcf7.consensus = dba.peakset(mcf7,mcf7$masks$Consensus,bRetrieve=TRUE)mcf7.consensus Boxplots for read count distributions within differentially bound sites dba.plotBox(DBA, contrast=1, method=DBA$config$AnalysisMethod, th=0.1, bUsePval=FALSE, bNormalized=TRUE,attribute=DBA_GROUP,bAll=FALSE, bAllIncreased=FALSE, bAllDecreased=FALSE,bDB=TRUE, bDBIncreased=TRUE, bDBDecreased=TRUE,pvalMethod=wilcox.test, number of contrast to use for boxplot.
method used for analysis (used in conjunction with contrast): significance threshold; all sites with FDR (or p-values, see bUsePval) less thanor equal to this value will be included in the boxplot.
logical indicating whether to use FDR (FALSE) or p-value (TRUE) for thresh-olding.
logical indicating that normalized data (using normalization factors computedby differential analysis method) should be plotted. FALSE uses raw count data.
attribute to use for determining groups of samples. Default (DBA_GROUP)plots the two groups used in the contrast. Possible values: • DBA_GROUP• DBA_ID• DBA_TISSUE• DBA_FACTOR• DBA_CONDITION• DBA_TREATMENT• DBA_REPLICATE• DBA_CONSENSUS• DBA_CALLER• DBA_CONTROL logical indicating if plot should include a set of boxplots using all counts, re-gardless of whether or not they pass the significance threshold.
logical indicating if plot should include a set of boxplots using all counts that in-crease in affinity, regardless of whether or not they pass the significance thresh-old.
logical indicating if plot should include a set of boxplots using all counts that de-crease in affinity, regardless of whether or not they pass the significance thresh-old.
logical indicating if plot should include a set of boxplots using all counts in sig-nificantly differentially bound sites (i.e. those that pass the significance thresh-old), regardless of whether they increase or decrease in affinity.
logical indicating if plot should include a set of boxplots using all counts insignificantly differentially bound sites that increase in affinity.
logical indicating if plot should include a set of boxplots using all counts insignificantly differentially bound sites that decrease in affinity.
method to use when computing matrix of p-values. If NULL, no matrix is com-puted, and NULL is returned; this may speed up processing if there are manyboxplots.
logical indicating if the default definition of positive affinity (higher affinity inthe second group of the contrast) should be reversed (i.e. positive affinity isdefined as being higher in the first group of the contrast).
vector of group numbers used to change the order that groups are plotted. IfNULL, default order is used (group order for DBA_GROUP, and the order theattribute values appear for other values of attribute).
vector of custom colors; if absent, default colors will be used.
Draws a boxplot showing distributions of read counts for various groups of samples under variousconditions. In default mode, draws six boxes: one pair of boxes showing the distribution of readcounts within all significantly differentially bound sites (one box for each sample group), one pairof boxes showing the distribution of read counts for significantly differentially bound sites thatincrease affinity in the second group, and a second pair of boxes showing the distribution of readcounts for significantly differentially bound sites that have higher mean affinity in the first group.
if pvalMethod is not NULL, returns a matrix of p-values indicating the significance of the differencebetween each pair of distributions.
#default boxplot includes all DB sites, then divided into those increasing# affinity in each groupdba.plotBox(tamoxifen) # plot non-normalized data for DB sites by tissue# (changing order to place Resistant samples last)dba.plotBox(tamoxifen, attribute=DBA_CONDITION, bDBIncreased=FALSE, bDBDecreased=FALSE, attribOrder=c(2,1), bNormalized=FALSE) dba.plotHeatmap(DBA, attributes=DBA$attributes, maxSites=1000, minval, maxval, contrast, method=DBA$config$AnalysisMethod,th=.1, bUsePval=FALSE, report, score,mask, sites, sortFun,correlations=TRUE, olPlot=DBA_COR,margin=10, colScheme="Greens", distMethod="pearson",.) attribute or vector of attributes to use for column labels: maximum number of binding sites to use in heatmap. Only used when not draw-ing a correlation heatmap (correlations=FALSE) Set all scores greater than this to maxval number of contrast to report on; if present, draws a heatmap based on a differ-ential binding affinity analysis (see dba.analyze). See dba.show(DBA, bCon-trast=T) to get contrast numbers.
analysis method (used in conjunction with contrast): significance threshold; all sites with FDR (or p-values, see bUsePval) less thanor equal to this value will be included in the report (subject to maxSites). Usedin conjunction with contrast.
logical indicating whether to use FDR (FALSE) or p-value (TRUE) for thresh-olding. Used in conjunction with contrast.
report (obtained from dba.report) specifying the data to be used . If this ispresent, the method, th, and bUsePval parameters are ignored. Used in con-junction with contrast.
Score to use for count data. Only used when plotting the global binding matrix(no contrast specified). One of: mask indicating a subset of peaksets to use when using global binding matrix(contrast is missing). See dba.mask.
logical vector indicating which sites to include; first maxSites of these. Onlyrelevant when using global binding matrix (contrast is missing).
function taking a vector of scores and returning a single value. Only relevantwhen using global binding matrix (contrast is missing). If present, the globalbinding matrix will be sorted (descending) on the results, and the first maxSitesused in the heatmap. Recommended sort function options include sd, mean,median, min.
logical indicating that a correlation heatmap should be plotted (TRUE). If FALSE,a binding heatmap of scores/reads is plotted. This parameter can also be set toa correlation record; see dba.overlap(mode=DBA_OLAP_ALL), in which casea correlation heatmap is plotted based on the specified correlation record, usingthe statistic specified in olPlot.
if correlations is specified as a dataframe returned by dba.overlap, indicateswhich statistic to plot. One of: • DBA_INALL number of peaks common to both samples Color scheme; see colorRampPalette RColorBrewer distance method for clustering; see Dist amap.
passed on to heatmap.2 (gplots), e.g. scale etc.
MODE: Correlation Heatmap plot using statistics for global binding matrix: dba.plotHeatmap(DBA, attributes=DBA$attributes, minval, maxval, correlations, olPlot, colScheme="Greens",distMethod="pearson", .) MODE: Correlation Heatmap plot using statistics for significantly differentially bound sites: dba.plotHeatmap(DBA, attributes=DBA$attributes, minval, maxval, contrast, method=DBA_EDGER,th=.1, bUsePval=F, overlaps, olPlot=DBA_COR, colScheme="Greens", distMethod="pearson", .) MODE: Binding heatmap plot using significantly differentially bound sites: dba.plotHeatmap(DBA, attributes, maxSites, minval, maxval, contrast, method, th, bUsePval, cor-relations=FALSE, colScheme, distMethod, .) MODE: Binding heatmap plot using the global binding matrix: dba.plotHeatmap(DBA, attributes, maxSites, minval, maxval, mask, sites, correlations=FALSE,sortFun, colScheme, distMethod, .) if correlations is not FALSE, the overlap/correlation matrix is returned.
data(tamoxifen_peaks)# peak overlap correlation heatmapdba.plotHeatmap(tamoxifen) data(tamoxifen_counts)# counts correlation heatmapdba.plotHeatmap(tamoxifen) data(tamoxifen_analysis)#correlation heatmap based on all normalized datadba.plotHeatmap(tamoxifen,contrast=1,th=1) #correlation heatmap based on DB sites onlydba.plotHeatmap(tamoxifen,contrast=1) #binding heatmap based on DB sitesdba.plotHeatmap(tamoxifen,contrast=1,correlations=FALSE) #binding heatmap based on 1,000 sites with highest variancedba.plotHeatmap(tamoxifen,contrast=1,th=1,correlations=FALSE,sortFun=var) Generate MA and scatter plots of differential binding analysis results Generates MA and scatter plots of differential binding analysis results.
dba.plotMA(DBA, contrast=1, method=DBA$config$AnalysisMethod, th=.1, bUsePval=FALSE,bNormalized=TRUE, factor="", bXY=FALSE, dotSize=.33, .) DBA object, on which dba.analyze should have been successfully run.
number of contrast to report on. See dba.show(DBA, bContrast=T) to get con-trast numbers.
method or vector of methods to plot results for: significance threshold; all sites with FDR (or p-values, see bUsePval) less thanor equal to this value will be colored red in the plot logical indicating whether to use FDR (FALSE) or p-value (TRUE) for thresh-olding.
logical indicating whether to plot normalized data using normalization factorscomputed by differential analysis method (TRUE) or raw read counts (FALSE).
string to be prepended to plot main title; e.g. factor name.
logical indicating whether to draw MA plot (FALSE) or XY scatter plot (TRUE).
#XY plots (with raw and normalized data)par(mfrow=c(1,2))dba.plotMA(tamoxifen,bXY=TRUE,bNormalized=FALSE)dba.plotMA(tamoxifen,bXY=TRUE,bNormalized=TRUE) dba.plotPCA(DBA, attributes, minval, maxval, contrast, method=DBA$config$AnalysisMethod,th=.1, bUsePval=FALSE, report, score,mask, sites, cor=FALSE,b3D=FALSE, vColors, dotSize, .) attribute or vector of attributes to use to color plotted points. Each unique com-bination of attribute values will be assigned a color. Chosen from: Note that DBA_GROUP is a special attribute which will result in samples fromeach group in a contrast being colored separately.
Set all scores greater than this to maxval number of contrast to use for PCA; if present, plots a PCA based on a differentialbinding affinity analysis (see dba.analyze). See dba.show(DBA, bContrast=T)to get contrast numbers. If missing, uses scores in the main binding matrix.
method used for analysis (used in conjunction with contrast): significance threshold; all sites with FDR (or p-values, see bUsePval) less thanor equal to this value will be included in the PCA, subject to maxVal. Used inconjunction with contrast.
if TRUE, uses p-value instead of FDR for thresholding. Used in conjunctionwith contrast.
report (obtained from dba.report) specifying the data to be used . If this ispresent, the method, th, and bUsePval parameters are ignored.
Score to use for count data. Only used when plotting the global binding matrix(no contrast specified). One of: mask indicating a subset of peaksets to use when using global binding matrix(contrast is missing). See dba.mask.
logical vector indicating which sites to include in PCA. Only relevant whenusing global binding matrix (contrast is missing).
a logical value indicating whether the calculation should use the correlation ma-trix or the covariance matrix. Passed into princomp.
logical indicating that three principal components should be plotted (requirespackage{rgl}). If FALSE, the first two principal components are plotted.
vector of custom colors; is absent, default colors will be used.
size of dots to plot; is absent, a default will be calculated.
arguments passed to plot or plot3d (rgl).
MODE: PCA plot using significantly differentially bound sites: dba.plotPCA(DBA, attributes, minval, maxval, contrast, method, th, bUsePval, b3D=F, vColors,dotSize, .) MODE: PCA plot using global binding matrix: dba.plotPCA(DBA, attributes, minval, maxval, mask, sites, b3D=F, vColors, dotSize, .) uses rgl package for 3D plots (if available) # peakcaller scores PCAdba.plotPCA(tamoxifen) # raw count correlation PCAdata(tamoxifen_analysis)dba.plotPCA(tamoxifen) #PCA based on normalized data for all sitesdba.plotPCA(tamoxifen,contrast=1,th=1) #PCA based on DB sites onlypar(mfrow=c(1,2))dba.plotPCA(tamoxifen,contrast=1)dba.plotPCA(tamoxifen,contrast=1,attributes=DBA_TISSUE) Draw 2-way or 3-way Venn diagrams of overlaps Draws 2-way or 3-way Venn diagrams of overlaps dba.plotVenn(DBA, mask, overlaps, label1, label2, label3, .) DBA object; if present, only the mask parameter will apply.
mask or vector of peakset numbers indicating which peaksets to include in Venndiagram. Only 2 or 3 peaksets should be included. See dba.mask. Only one ofmask or overlaps is used.
overlap record, as computed by dba.overlap(Report=DBA_OLAP_PEAKS). Onlyone of mask or overlaps is used.
arguments passed on to vennDiagram{limma} par(mfrow=c(2,2))# 2-way Venndba.plotVenn(tamoxifen,6:7)dba.plotVenn(tamoxifen,tamoxifen$masks$ZR75) # 3-way Venn (done two different ways)dba.plotVenn(tamoxifen,tamoxifen$masks$MCF7&tamoxifen$masks$Responsive)olaps = dba.overlap(tamoxifen,tamoxifen$masks$MCF7&tamoxifen$masks$Responsive)dba.plotVenn(tamoxifen,overlaps=olaps, label1="Rep 1",label2="Rep 2",label3="Rep 3",main="MCF7 (Responsive) Replicates") #Venn of overlapsResponsive=dba(tamoxifen,tamoxifen$masks$Responsive)ResponsiveResponsive = dba.peakset(Responsive,1:3,sampID="MCF7")Responsive = dba.peakset(Responsive,4:5,sampID="T47D")Responsive = dba.peakset(Responsive,6:7,sampID="ZR75")dba.plotVenn(Responsive,Responsive$masks$Consensus) Generate a report for a differential binding affinity analysis Generates a report for a differential binding affinity analysis dba.report(DBA, contrast=1, method=DBA$config$AnalysisMethod, bCalled=FALSE, bCounts=FALSE, bCalledDetail=FALSE,file,initString=DBA$config$reportInit,ext=’csv’,DataType=DBA$config$DataType) DBA object. A differential binding affinity analysis needs to have been previ-ously carried out (see dba.analyze).
contrast number to report on. See dba.show(DBA, bContrast=T) to get contrastnumbers.
significance threshold; all sites with FDR (or p-values, see bUsePval) less thanor equal to this value will be included in the report. A value of 1 will include allbinding sites in the report.
logical indicating whether to use FDR (FALSE) or p-value (TRUE) for thresh-olding.
only sites with an absolute Fold value greater than equal to this will be includedin the report.
logical indicating that normalized data (using normalization factors computedby differential analysis method) should be reported. FALSE uses raw countdata.
logical indicating that peak caller status should be included (if available from aprevious call to dba.count(bCalledMasks=TRUE)). This will add a column foreach group, each indicating the number of samples in the group identified as apeak in the original peaksets.
logical indicating that count data for individual samples should be reported aswell as group statistics. Columns are added for each sample in the first group,followed by columns for each sample in the second group.
logical indicating that peak caller status should be included for each sample (ifavailable). Columns are added for each sample in the first group, followed bycolumns for each sample in the second group.
if present, also save the report to a comma separated value (csv) file, using thisfilename.
if saving to a file, pre-pend this string to the filename.
if saving to a file, append this extension to the filename.
Can be set as default behavior by setting DBA$config$DataType.
A report dataframe or RangedData object, with a row for each binding site within the thresholdingparameters, and the following columns: Concentration – mean (log) reads across all samples in both groups Group 1 Concentration – mean (log) reads across all samples first group Group 2 Concentration – mean (log) reads across all samples in second group Fold difference – mean fold difference of binding affinity of group 1 over group2 (Conc1 - Conc2). Absolute value indicates magnitude of the difference, andsign indicates which one is bound with higher affinity, with a positive valueindicating higher affinity in the first group p-value calculation – statistic indicating significance of difference (likelihooddifference is not attributable to chance) adjusted p-value calculation – p-value subjected to multiple-testing correction If bCalled is TRUE and caller status is available, two more columns will follow: Number of samples in group 1 that identified this binding site as a peak Number of samples in group 2 that identified this binding site as a peak If bCounts is TRUE, a column will be present for each sample in group 1, followed by each samplein group 2. The Sample ID will be used as the column header. This column contains the read countsfor the sample.
If bCalledDetail is TRUE, a column will be present for each sample in group 1, followed by eachsample in group 2. The Sample ID will be used as the column header. This column contains a "+"to indicate for which sites the sample was called as a peak, and a "-" if it was not so identified.
tamoxifen.DB = dba.report(tamoxifen)tamoxifen.DB tamoxifen.DB = dba.report(tamoxifen,th=.05,bUsePval=TRUE,fold=2)tamoxifen.DB #Retrieve all sites with confidence stats# and how many times each site was identified as a peaktamoxifen.DB = dba.report(tamoxifen, th=1, bCalled=TRUE)tamoxifen.DB #Retrieve all sites with confidence stats and normalized counts tamoxifen.DB = dba.report(tamoxifen,th=1,bCounts=TRUE)tamoxifen.DB #Retrieve all sites with confidence stats and raw countstamoxifen.DB = dba.report(tamoxifen,th=1,bCounts=TRUE,bNormalized=FALSE)tamoxifen.DB dba.save(DBA, file=’DBA’, dir=’.’, pre=’dba_’, ext=’RData’, bMinimize=FALSE) logical indicating saved DBA object should be compressed as much as possible.
string containing full path and filename.
data(tamoxifen_peaks)savefile = dba.save(tamoxifen,’tamoxifenPeaks’)savefiletamoxifen = dba.load(’tamoxifenPeaks’)unlink(savefile) List attributes of peaksets of contrasts associated with a DBA object Returns attributes of peaksets and/or contrasts associated with a DBA object.
dba.show(DBA, mask, attributes, bContrasts=FALSE, th=0.1, bUsePval=FALSE) mask of peaksets for which to get attributes (used when obtaining peakset at-tributes, i.e. bContrasts=FALSE).
attribute or vector of attributes to retrieve. Number of intervals is always shown.
Used when obtaining peakset attributes, i.e. bContrasts=FALSE. Values: logical indicating whether peaksets or contrast attributes are to be retrieved.
TRUE retrieves a dataframe of contrast information instead of peakset attributes.
If no contrasts are set, returns possible contrasts. See dba.contrast.
if bContrasts is TRUE, then th is used as the threshold for determining howmany significant sites there are for each contrast. Only relevant when obtainingcontrast attributes (bContrasts=TRUE) and dba.analyze has been run.
logical indicating that p-values will be used (along with th) to determine howmany significant sites there are for each contrast; if FALSE, adjusted p-values(FDR) are used. Only relevant when obtaining contrast attributes (bContrasts=TRUE)and dba.analyze has been run.
MODE: Return attributes of peaksets associated with a DBA object: MODE: Return contrasts associated with a DBA object: DiffBind – DBA global constant variables If bContrasts == FALSE, each row represents a peakset, and each column is an attributes, with thefinal column, Intervals, indicating how many sites there are in the peakset.
If bContrasts == TRUE, each row represent a contrast, with the following columns: Number of samples in first group of contrast Number of samples in first group of contrast if dba.analyze has been successfully run, there there will be up to two more columns showing thenumber of significant differentially bound (DB) sites identified for Number of significantly differentially bound sites identified using edgeR Number of significantly differentially bound sites identified using DESeq data(tamoxifen_peaks)dba.show(tamoxifen)dba.show(tamoxifen,tamoxifen$masks$Responsive)dba.show(tamoxifen,attributes=c(DBA_TISSUE,DBA_REPLICATE,DBA_CONDITION)) data(tamoxifen_counts)tamoxifen = dba.contrast(tamoxifen)dba.show(tamoxifen,bContrasts=TRUE) DiffBind -- DBA global constant variables Constant variables used in DiffBind package Constant variables used in DiffBind package DBA_IDDBA_FACTORDBA_TISSUEDBA_CONDITIONDBA_TREATMENTDBA_REPLICATE DiffBind – DBA global constant variables DBA_SCORE_READSDBA_SCORE_READS_MINUSDBA_SCORE_READS_FOLDDBA_SCORE_RPKMDBA_SCORE_RPKM_FOLDDBA_SCORE_TMM_READS_FULLDBA_SCORE_TMM_READS_EFFECTIVEDBA_SCORE_TMM_MINUS_FULLDBA_SCORE_TMM_MINUS_EFFECTIVE DBA_EDGERDBA_DESEQDBA_EDGER_BLOCKDBA_DESEQ_BLOCKDBA_EDGER_CLASSICDBA_DESEQ_CLASSICDBA_EDGER_GLMDBA_DESEQ_GLM DBA_DATA_FRAMEDBA_DATA_GRANGESDBA_DATA_RANGEDDATA DBA peakset metadata: Is this a consensus peakset? DBA peakset metadata: ID of Control sample DBA peakset metadata: color PCA plot using contras groups DBA_OLAP_PEAKS dba.overlap mode: return overlapping/unique peaksets dba.overlap mode: return report of correlations/overlaps for each pair of samples DiffBind – DBA global constant variables dba.count score is number of reads in ChIP dba.count score is number of reads in ChIP divided by number of reads in Con-trol dba.count score is number of reads in ChIP minus number of reads in Control DBA_SCORE_RPKM dba.count score is RPKM of ChIPDBA_SCORE_RPKM_FOLD dba.count score is RPKM of ChIP divided by RPKM of Control dba.count score is TMM normalized (using edgeR), using ChIP read counts andFull Library size dba.count score is TMM normalized (using edgeR), using ChIP read counts andEffective Library size dba.count score is TMM normalized (using edgeR), using ChIP read counts mi-nus Control read counts and Full Library size dba.count score is TMM normalized (using edgeR), using ChIP read counts mi-nus Control read counts and Effective Library size differential analysis method: edgeR (default: DBA_EDGER_GLM) differential analysis method: DESeq (default: DBA_DESEQ_CLASSIC) differential analysis method: "classic" edgeR for two-group comparisons differential analysis method: "classic" DESeq for two-group comparisons differential analysis method: use GLM in edgeR for two-group comparisons differential analysis method: use GLM in DESeq for two-group comparisons differential analysis method: edgeR with blocking factors (GLM) differential analysis method: DESeq with blocking factors (GLM) Use GRanges class for peaksets and reports. This is the default (DBA$config$DataType= DBA_DATA_GRANGES).
Use RangedData class for peaksets and reports. Can be set as default (DBA$config$DataType= DBA_DATA_RANGEDDATA).
DBA_DATA_FRAME Use data.frame class for peaksets and reports. Can be set as default (DBA$config$DataType Variables with ALL CAP names are used as constants within DiffBind.
DiffBind (DiffBind-package), DiffBind -- DBA global constant plot.DBA (DBA object methods), print.DBA (DBA object methods),

Source: ftp://bne-a-ms5.retain.aarnet.edu.au/pub/bioconductor/packages/2.10/bioc/manuals/DiffBind/man/DiffBind.pdf