FRASER can not identify known abnormal splicing event

[qj009]

Hello,

I am running FRASER (version: 2.2.0.) to identify abnormal splicing events in my data. However, it cannot identify the known differential expressed splicing events we have. The numbers are as listed below:

The adjusted P value is 1 and P value is a little bit smaller than 0.05.
The total number of events in the test with several read count filters I tried are listed below:

• default: at least one sample has 20 or more reads, and 25% or more of the samples have at least 10 reads;
• Others:
  at least one sample with 10 reads;
  at least one sample has 20 or more reads, and 5% or more of the samples have at least 1 read;

The script example I used is as below:

# creat fraser object 
settings <- FraserDataSet(colData=sampleTable, workingDir=study) 
# count reads 
fds <- countRNAData(settings, minExpressionInOneSample=0, BPPARAM=MulticoreParam(nthread), filter=TRUE) 
# Calculate PSI value 
fds <- calculatePSIValues(fds) 
# filter 
# quantile: tried 0.75 and 0.95 
# RCquantile: tried 0,1,10 
# RCone: tried 10,20 
fds <- filterExpressionAndVariability(fds, minDeltaPsi=0, quantile=quantile, quantileMinExpression = RCquantile, minExpressionInOneSample = RCone, filter=FALSE,BPPARAM=MulticoreParam(nthread)) 
fds <- fds[mcols(fds, type="j")[,"passed"]] 
# run FRASER 
fds <- FRASER(fds,BPPARAM=MulticoreParam(nthread)) 
# output result 
# Usually did not set any threshold for padjCutoff and deltaPsiCutoff to to look up result of positive control events 
res <- results(fds, padjCutoff = padjust, deltaPsiCutoff = deltapsi, BPPARAM=MulticoreParam(nthread)) 

With the filter : at least one sample has 10 or more reads, the adjusted P value of known events in the target sample is 1 and P value is 0.008. There is over 50 samples included in this analysis. The known events was filtered out if using filter: at least one sample has 20 or more reads.

Any suggestions are appreciated. Thank you!

[vyepez88]

Hi, it can be the same problem you faced in OUTRIDER. FRASER is meant for aberrant, not alternative splicing detection. What is your expected change in splicing? Can you see it in the raw data?
Alternatively, try to extract all the results results(fds, all = T) and check it.

[qj009]

Hello, thank you so much for your previously reply and advice in both FRASER and OUTRIDER. Both were helpful.

I had a exon skipping splicing events with relative extreme situation. It can be seen clearly in one sample (sample A) BAM file, FRASER calculate psi3 is 0.41 and psi5 is 0.36 (did not use jaccard since there were DNA contaminations in my samples, only want to include junction reads), but not in other sample BAMs (supporting reads for the skipped exon junction were 0 in other samples). However, FRASER calculated deltapsi is 0.03 for sample A which means the difference for this SE event between sample A and other samples were minimal which should not be true.

I wonder is this situation (one splice event highly expressed in one sample, but no expression in rest of 50 samples) break the assumption of FRASER? Thank you again for your help!