User:Lindenb/Notebook/UMR915/20101201

Backup
Matthieu disabled Bacula

Belgium
working on belgium data

Dindel
From SLS about Dindel: Here is what we do by default - we now use the latest version of dindel (1.01) and we call candidates per sample for each sample in a project. We merge all the candidates together into a project set and then restrict them to the PCCR regions (union of baits and targets) +/- 100 bp (we don't apply any other filters to the candidates). We don't do anything fancy like adding 1000G or dbSNP indels to the candidate list as known indels are usually of less interest to us anyway. We then run dindel on a per sample basis using the merged candidate list. The final output is dindels VCF

v4 with consequences (via ensembl pipeline code) and known indels (rough matching e.g. indel +/- 10bp) annotated.

We can also run across all samples at once (as a pool) which is quicker, but this is only really recommended by Kees for low depth data.