Difference of AT/GC dropout metric explanation between commands
The explanations for AT/GC Dropout metrics are different between commands.
Please look at the explanation of the metrics in the following tools
GcBiasSummaryMetrics - AT_DROPOUT/GC_DROPOUT
(%ref_at_gc - %reads_at_gc) and summing all positive values
HsMetrics/TargetedPcrMetrics - AT_DROPOUT/GC_DROPOUT
a = fraction of target territory, and b = fraction of aligned reads
abs(sum(a-b when a-b < 0))
In GcBiasSummaryMetrics, it says that picard sums up all the positives.
But in HsMetrics and TargetedPcrMetrics, it says that picard sums up all the negatives.
I don't understand why they have to be different.
I understand the explantation in GcBiasSummaryMetrics.
If the percentage of aligned read is lower compared to the relative lengths of low GC regions, then we can say the low GC regions are depleted of reads. This is AT dropout.
But for HsMetrics and TargetedPcrMetrics,
If a(fraction of target territory) - b(fraction of aligned reads) is negative,
then b(fraction of aligned reads) is larger and reads are enriched in this region.
This doesn't seem to be a correct measure of "Dropout."
Could you explain why, for targeted sequencing, the AT/GC dropouts require the opposite way of calculation?
-
Hi JYoung Lee
Those are 2 separate metrics (There are no distinct tools with those names.) that report AT_DROPOUT and GC_DROPOUT for different purposes.
HsMetrics and TargettedPcrMetrics report these values as how each sample is biased their AT and GC rich regions are in terms of coverage and readcount.
GcBiasSummaryMetrics report these values as how each lane in an sequencer is different in terms of AT and GC content within.
Using similar naming is what is really causing the confusion but don't worry. They are for 2 different purposes and yet are not comparable.
I hope this helps.
-
Thank you for your explanation, Gökalp Çelik.
Then would it be correct to understand them as follows?
1)
GC/AT Dropouts for GcBiasSummaryMetrics measure depletion of reads in high/low GC regions.2-1)
GC/AT Dropouts for HsMetric and TargetedPcrMetrics measure how narrow the covered region( with aligned read >1) is, compared to the number of reads aligned in that region.2-2)
and "a" means the followinga : proportion of regions that are of certain GC content and covered by at least 1 read relative to total target territory
-
For 1) this measure is per lane of a flowcell.
For 2) Exactly.
-
Thank you for your clarification!
Please sign in to leave a comment.
4 comments