Is Delly calling SV sample-by-sample?

[Jerry Xu]

I usually ran 10 low-coverage whole-genome sequences together using Delly with parallel computing.
/delly -g g1k37.fa -t DEL sample1.bam sample2.bam sample3.bam...

And I thought Delly retrieves read-pair/split-read information and call SV on a single sample, then merge per-sample SV callings into multi-sample vcf file. Is that correct?

But today I tried to run those same low-coverage whole-genome sequences one by one,
delly -g g1k37.fa -t DEL sample1.bam
delly -g g1k37.fa -t DEL sample2.bam
delly -g g1k37.fa -t DEL sample3.bam

and found the deletion calls (~2000) are quite fewer than running in parallel (~7000) as before. This makes me feel suspicious maybe Delly works by combining different samples to increase sensitivity when multiple samples are submitted to run together?

Any comments?

Many thanks!

[Tobias Rausch]

Delly pools the SV evidence from all input samples so for low-coverage samples it's indeed better to run multiple samples together to increase sensitivity.

-Tobias

--
You received this message because you are subscribed to the Google Groups "delly-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to delly-users...@googlegroups.com.
To post to this group, send email to delly...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[John Farrell]

When the bam files have 30x coverage, is running on multiple samples also useful?  Is there a maximum number of samples that would be recommended to be run together?

John

[Tobias Rausch]

Hi John,

At high coverage (>=30x) it's fine to run Delly sample-by-sample for germline SVs and in a paired tumor-normal fashion for somatic SVs. Just make sure that you afterwards merge, re-genotype and filter the SVs as described in the README file.

Best, Tobias