the recommended sequencing depth
Closed this issue · 2 comments
Hi,
I am wondering about the recommended sequencing depth for SVIM to detect correct SV.
Hi Charlie,
well, higher sequencing depth is of course better. But it is always a trade-off between sequencing cost and the calling accuracy you need to reach. I uploaded a plot below that might help you with the decision.
It shows F1 score (harmonic mean between precision and recall) reached by SVIM on the HG002 individual. I used three very recent long-read datasets (PacBio CLR, PacBio CCS, ONT) and measured SVIM's performance using the GIAB SV benchmark set. For each dataset, 10 subsamples of 10%, 20%, ..., 90% and 100% were produced. As you can see, F1 score (y-axis) increases with increasing sequencing depth (x-axis) until saturation is reached around ~20x.
I hope that helps.
Cheers
David
thank you very much!!!