movAPAswitch: Analyze APA site switching.
In BMILAB/movAPA: movAPA: Modeling and Visualization of Dynamics of Alternative PolyAdenylation

movAPAswitch

R Documentation

Analyze APA site switching.

Description

movAPAswitch analyzes APA site switching for genes with both canonical and non-canonical APA sites. It performs analyses for all condition pairs of the given group. If you only need a specific condition pair, call subsetPACds first to filter used conditions.

Usage

movAPAswitch(
  PACds,
  group,
  avgPACtag = 5,
  avgGeneTag = 10,
  only3UTR = FALSE,
  mergeReps = "pool",
  aMovDEPACRes = NULL,
  DEPAC.padjThd = NULL,
  nDEPAC = NULL,
  mindist = 50,
  fisherThd = 0.1,
  logFCThd = 1,
  cross = FALSE,
  selectOne = NULL
)

Arguments

`PACds`	a PACdataset.
`group`	a sample group name to get all conditions under the group.
`avgPACtag`	filter PACs with average tag number between cond1-cond2 >= avgPACtag.
`avgGeneTag`	the same as avgPACtag but for gene.
`only3UTR`	If TRUE, then first filter 3UTR PACs.
`mergeReps`	pool or avg replicates of the same condition, default is pool.
`aMovDEPACRes`	an object of movDEPACRes to get DE PAC results. If nDEPAC>0, should provide.
`DEPAC.padjThd`	to filter DE PACs with padj<DEPAC.padjThd.
`nDEPAC`	can be NULL (0), 1 or 2, denoting that the switching pair has 0, 1 or 2 DE PACs. if nDEPAC>0 then should provide aMovDEPACRes.
`mindist`	the min distance between two PACs to be considered for switching analyses.
`fisherThd`	used to filter DE PAC list from DE/DEX method by pvalue<fisherThd.
`logFCThd`	used to filter \|rcij\| >=logFCThd, where \|rcij\| is the log2(ratio of PA1 between two samples/ratio of PA2).
`cross`	Whether the change of ratio should be crossing for PA1 and PA2.
`selectOne`	can be 'farest','logFC','fisherPV' to output only one switching pair if a gene has multiple switching pairs.

Details

This function is similar to movUTRtrend(method=DE/DEX), but it can be used for non-3'UTR PACs. If nDEPAC>0, then first get DE PACs from aMovDEPACRes. For genes with more than two PACs, each pair of PACs with distance>=mindist is computed for switching. If a pair of PACs meets the following criteria, then it is switching: (1) DEPAC# (by DEPAC.padjThd) >= nDEPAC; (2) Pvalue of the Fisher's text of the two PACs between conditions < fisherThd; (3) |Relative usage (RU)| of the two PACs between conditions >=logFCThd; Given PAi and PAj between samples A and B, RU=log2(ejB/ejA)-log2(eiB/eiA) (4) If cross=T, then expression levels of the two PACs are switched between conditions.

Value

An object of movAPASwitchRes similar to movUTRTrendRes.

Examples

# Detect 3UTR APA switching using DEXseq, which is similar to movUTRtrend(method=DEX)
data(PACds)
## First get DE PAC results by DEXseq.
DEXPAC=movDEPAC(PACds, method='DEXseq', group='group', minSumPAT=10)
## Then get 3UTR switching genes, usig selectOne=NULL to detect all pairs of switching PACs.
swDEX=movAPAswitch(PACds, group='group',aMovDEPACRes=DEXPAC,
                   avgPACtag=5, avgGeneTag=10,
                   only3UTR=TRUE,
                   DEPAC.padjThd=0.1, nDEPAC=1,
                   mindist=50, fisherThd=0.1, logFCThd=0.5, cross=FALSE, selectOne=NULL)

lapply(swDEX@fullList,nrow)
head(swDEX@fullList[["anther.embryo"]])
head(swDEX@pairwise$padj

## Detect APA switching events involving non-3UTR PACs.
## First get DE PAC results by DESeq2 (or DEXseq).
DEXPAC=movDEPAC(PACds, method='DEXseq', group='group', minSumPAT=10)
## Then detect APA switching events,
## using selectOne=NULL to get all pairs of switching PACs.
swDE=movAPAswitch(PACds, group='group',aMovDEPACRes=DEXPAC,
                 avgPACtag=10, avgGeneTag=20,
                 only3UTR=FALSE,
                 DEPAC.padjThd=0.1, nDEPAC=1,
                 mindist=50, fisherThd=0.1, logFCThd=0.5,
                 cross=FALSE, selectOne=NULL)
## Stat the switching results.
stat=movStat(object=swDE, padjThd=0.1, valueThd=1)
stat$nsig

BMILAB/movAPA documentation built on Jan. 3, 2024, 11:09 p.m.