defintion of extremely low coverage

Question

[statfa]

hi,

I have searched for the answer but I still need more clear explanation to understand the concept.

we need to filter out exons that have extremely low coverage (median read depth across all samples less than 20, which mostly reflect capture failure).

As I know, the concept of coverage depth is different from read depth... coverage depth simply means how much each segmentation is covered by the sequencing platform. While read depth means how many times a segmentation has been read. (please correct me if I'm wrong).

now my question is how can "extremely low coverage" be defined as "median read depth across all samples less than 20"?

could someone give me a clear explanation of "median" read depth?

Thank you so much

Answer 1

[Devon Ryan]

1. Count the number of reads per region in each sample.
   • Be sure to filter out anything that's excluded by your variant caller.
2. Take the median per region across samples
3. Filter

[statfa]

Thank you so much... I appreciate that.

Answer 2

[igor]

For most people, coverage and read depth are the same. If a certain nucleotide has 10 reads overlapping it, then the coverage for it is 10X. A region of interest (a single exon or an entire exome) has multiple nucleotides. Each nucleotide has a specific coverage value. You can take a median of those to get the median coverage.

Also, you can use something like GATK DepthOfCoverage to do the calculation for you: https://www.broadinstitute.org/gatk/guide/tooldocs/org_broadinstitute_gatk_tools_walkers_coverage_DepthOfCoverage.php

[John]

I always thought "depth" meant the average number of reads above any given base, and "coverage" meant the amount of genome with at least 1 read covering it..?

[Devon Ryan]

That can be the case, but isn't always. It's a lovely ambiguous language we speak.

[John]

Yes I suppose "number of times the base is covered by reads" is also depth. Perhaps we should all switch to unambiguous emojis for technical talk. Like:

"All technical 🐑🐑 were 🔬 and found to have a ♨️ and 🍉 leading us to 🤔 that 👾 🛀 🎱 🎻."

Just kidding, that makes no sense - but only because the Unicode Consortium hasn't accepted an application for read depth and genomic coverage emojis... yet.

[Devon Ryan]

Then we'd have problems like just now, where I can only see one of the emojis... :P

[statfa]

Thank you so much for the information... I appreciate that.