Using FRASER on sample subsets to calculate p-value

[irikas]

Hi,
I have a FraserDataSet that looks similar to the following.

-------------------- Sample data table -----------------
# A tibble: 22 × 6
   sampleID bamFile      group gene   pairedEnd SeqLevelStyle
   <chr>    <chr>        <int> <chr>  <lgl>     <chr>        
 1 CON_1    CON_1.bam        1 NA      TRUE      UCSC         
 2 CON_2    CON_2.bam        1 NA      TRUE      UCSC         
 3 Trt1_1   Trt1_1.bam       2 NA      TRUE      UCSC         
 4 Trt1_2   Trt1_2.bam       2 NA      TRUE      UCSC
 5 Trt2_1   Trt2_1.bam       3 NA      TRUE      UCSC         
 6 Trt2_2   Trt2_2.bam       3 NA      TRUE      UCSC
 7 Trt1.2_1 Trt1.2_1.bam     4 NA      TRUE      UCSC         
 8 Trt1.2_2 Trt1.2_2.bam     4 NA      TRUE      UCSC          

Essentially there are multiple samples in the same group and I would like to take the average PSI value of each junction in each group to compare the groups to each other. Additionally, I would like to compare each treatment group to the control group separately and obtain PSI values that way. Is that possible? Would I simply need to create multiple FRASER objects?

Finally, due to having a variety of negative controls, I would like to (essentially) subtract the PSI of samples 5&6 from 7&8 to determine the contribution that a treatment has on its own. Is that possible with the results table?

Thank you!

EDIT: I did see issue #26 and am considering also using LeafCutter

[vyepez88]

Hi, as you mention in your 'edit', indeed it seems that you need a differential splicing tool instead. However, you can extract the PSI from the samples. Instead of from the results table, you have to do it from the FRASER DataSet (fds) object. After loading it, check all the assays with sort(names(assays(fds))). The assays contain the split counts, psi, pvalue, etc. In addition, you will need the genomic ranges that you can obtain using granges(fds). Good luck!