Recommended thresholds for filtering modkit pileup bedMethyl output (Nmod, fraction_modified, Nfail)
Hi all,
I'm working with ONT methylation data and using modkit pileup to generate a bedMethyl output for CpG methylation (5mC) using the following command:
modkit pileup input.bam output.bed.gz \
--modified-bases 5mC \
--cpg \
--reference genome.fasta
I want to filter confidently methylated positions but I’m unsure what thresholds are best to use.
Can anyone clarify:
- Are there recommended default thresholds (especially for 5mC) for confident sites?
- Is there any biological or technical justification (e.g. coverage, confidence) for choosing values like 5 or 70%?
- How do thresholds like
--filter-thresholdor--filter-percentileinteract with Nfail?
Thanks everyone in advance!!
• 1,033 views
•
link
updated
by
GenoMax
•
written
by
alexandrakortsi •
•
written
by
alexandrakortsi •
0 answers
No answers yet.
Log in to answer this question.
Question #3 is addressed partly here: https://nanoporetech.github.io/modkit/faq.html#what-value-for---filter-threshold-should-i-use
If you have not had a look at the documentation you may want to do that while you wait for answers.
Thank you very much! I understand modkit uses certain thresholds, but shouldd we accept everything that passes these thresholds? Does it matter on how many reads this modification was found ? And if so, how?
ONT has likely coded in defaults that should work for most use cases. If you are working with an edge case you could email the support and ask.