nuno-agostinho
commented
1 year ago Hey @ryanyord, sorry for the delay in replying!
By default, psichomics automatically filters alternative splicing events with a minimum of 10 junction read counts supporting either inclusion or exclusion of the alternative sequence, avoiding events with insufficient evidence. However, you can further filter events with psichomics (or any other tool of your choice), as desired.
Quality control is dependant on the data: you will need to visualise/plot your data so you can answer those questions. As a reference, in our methods chapter (Interactive Alternative Splicing Analysis of Human Stem Cells Using psichomics), we used psichomics to filter alternative splicing events using all of the metrics displayed in the figure and table below. If needed, I would be glad to email you the PDF of our book chapter on psichomics, simply send me your email.
| Metric | Description | Example |
|---|---|---|
| Median PSI | Filter out low-variance AS events based on median PSI, as well as avoid events whose median PSI is near 0 and 1 (i.e. mostly constitutive events) | Median PSI between 0.05 and 0.95 |
| PSI variance | Events that vary across samples based on PSI variance | log10(PSI variance) > −3 |
| PSI range | Further filter events based on their PSI range, as a surrogate for the minimum changes in AS that can be considered biologically meaningful | PSI range > 0.15. |
Again, your data shall dictate which thresholds you use, but I hope you get the gist.
Please feel free to contact me in case you have any doubts. Thank you!
Cheers, Nuno
ryanyord
Hello,
I've been looking through documentation searching for any advice/suggestions on normalization or Quality Control.
Specifically when using quantifySplicing() function using user-uploaded data I was wondering if there was any suggestions for QC prior to input? After attaining PSI values is there any suggestions on QC/filtering the data?
Thank you