Dindel is seeing too many reads, potential bug?

[Heisman]

Hey all,

I have just figured out how to download and use dindel and I am trying to compare it to samtools mpileup. Most of the calls are the same. However, when looking at a dindel call in IGV, I noticed that there is only 1 read covering that position. Yet, dindel gave the following output line in the VCF file:

chr19   11243209        .       c       cG      128     PASS    DP=12;NF=0;NR=4;NRS=3;NFS=1;HP=1        GT:GQ   1/1:12

When looking at the depthofcoverage file, it states that there is only 1 read. Yet, dindel sees 4 reads. Does anybody have any idea why this could happen? I've attached a screenshot of this position in IGV.

[ratope]

I had the same doubt. After checking the intermediate results of Dindel, I realized that DP value in the VCF is the number of reads that cover the window that is processed by Dindel.

For example, in my case the VCF contains an indel in the position 6680:

#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	SAMPLE
chromosome_II	6680	.	TATA	T	118	PASS	DP=230;NF=0;NR=3;NRS=13;NFS=23;HP=1	GT:GQ	0/1:118

The "depth" is:

• DP=230 according to Dindel (INFO column in the VCF file), but

• depth=45 according to IGV (in fact, according to the pileup file).

The information displayed by the "step 3" of Dindel showed this:

(...)
****
 tid: chromosome_II [B]pos: 6681 leftPos: 6620  rightPos: 6742[/B]
Fetching reads....
[B]Number of reads: 230[/B] out of 77463 # unmapped reads: 0 numReadsUnknownLib: 0 numChrMismatch: 0 numMappedWithoutMate: 2 numUnmappedWithoutMate: 0
candidate_var@pos: 6681 6680,-ATA
aligned_var@pos 6681 6656 A=>G
aligned_var@pos 6681 6657 T=>A
aligned_var@pos 6681 6680 -ATA
[empiricalDistributionMethod] Number of haplotypes: 8
Filtered 0 haplotypes.
ll_ref: -1085.49 max_ll_indel: -1058.3 qual: 118.099
(...)

My interpretation is that DP is the number of reads covering the positions (window) 6620-6742, and not only those covering the "starting" point of the indel (6680).

Hope it is useful.

[ratope]

More simple!

My suspicion was correct, but there is a more straightforward way to confirm it.

From the header of the VCF file produced by Dindel:

##INFO=<ID=[B]DP[/B],Number=1,Type=Integer,Description="[B]Total number of reads in haplotype window[/B]">