-
Notifications
You must be signed in to change notification settings - Fork 67
VAF Cutoff Filter Experiment for SNV callers #173
VAF Cutoff Filter Experiment for SNV callers #173
Conversation
…cansav09/snv_calculations
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, I just have a few comments below.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Overall, looks good and gets the point across. I was confused by the variance calculations, and I am not sure whether they really are useful in this context. It isn't really a sampling question (you chose the cutoffs; they aren't random sets), so the changes in TMB aren't really something that variance makes sense for. I'd probably just eliminate that section.
The main point I was searching for here was "how much are individual sample's TMB's changing as we change VAF cutoffs?" I get that variance is not quite appropriate for this, but wasn't sure how to show this. That being said, I am fine with dropping that whole section and moving on. |
Okay, @cbethell and @jashapiro, I've removed the confusing sample TMB variance section, added a summary at the top of the file, but did not make a separate |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done and dusted
Purpose/implementation
This is a side analysis that originated from this original PR and analysis.
Note: This PR depends on changes made in #172 and shouldn't be reviewed or tested until that is merged.
After doing some initial comparisons of the callers TMB stats and finding them not well correlated to each other at all; the question from @jashapiro arose of whether increasing the VAF filter cutoff would help make the TMBs be more consistent across samples. Hence this analysis was born.
Issue
For SNV caller comparison #161, #103 and Tumor Mutation Burden #3 and sort of #11
Directions for reviewers
See the rough draft of this notebook output from this notebook here: https://cansavvy.github.io/openpbta-notebook-concept/snv-callers/vaf_cutoff_experiment.nb.html
Some questions for this PR:
What do you think of the current analyses?
Are there other analyses you would like to see added?
Should I put the custom functions in their own script in
util
folder as has been done for the previous functions in this analysis folder?Results
See the rough draft of this notebook output from this notebook here: https://cansavvy.github.io/openpbta-notebook-concept/snv-callers/vaf_cutoff_experiment.nb.html
Preliminary conclusion:
It looks like increasing the VAF cutoff improves the callers' TMB calculations in how much they are related to each other. However, note that although VarDicts TMBs get recovered a bit by increasing a VAF filter cutoff, VarDict's TMBs are still the least associated with the other callers'.
Docker and continuous integration
Main RPackages added to the Dockerfile:
PR Checklist