[Samtools-help] bug in samtools mpileup?

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hi,

I’ve found something very strange with samtools mpileup, looks like a bug.

If I provide the reference fasta I get completely different results than if I don’t.

Look at this example:
./samtools-1.3.1/samtools mpileup -r 20:36052981 -f human_g1k_v37.fasta file.bam
20	36052981	A	2	,G	>1
20	36052982	G	2	,.	>1
20	36052983	G	2	,$.	<3
20	36052984	A	9	.,..,,.,.	<B7C77779
20	36052985	G	9	.,..,,.,.	<B7C77779
20	36052986	A	9	.,..$,,.,.	<F7=77779
20	36052987	A	8	GgGggGgG	</777779
20	36052988	G	8	.,.,,.,.	<F777779
20	36052989	G	8	.,.,,.,.	<F777779
20	36052990	A	8	.,.,,.,.	FFKKKKKK
20	36052991	G	8	.,.,,.,.	FFKKKKKK
20	36052992	C	8	.,.,,.,.	FFKKKKKK
20	36052993	T	8	.,.,,.,.	FFKKKKKK
20	36052994	G	8	.,.,,.,.	FFKKKKKK
20	36052995	A	8	.,.,,.,.	BFKKKKKK
20	36052996	G	8	.,.$,,.,.	BF>KKKKK

Now look what we get if we do not provide a reference:
./samtools-1.3.1/samtools mpileup -r 20:36052981 file.bam
20	36052981	N	11	aaGgGGggGgG	BBFFKKKKKKK
20	36052982	N	11	g$gGgGGggGgG	BBFFKKKKKKK
20	36052983	N	10	g$GgGGggGgG	<FFKKKKKKK
20	36052984	N	9	AaAAaaAaA	FBKKKKKKK
20	36052985	N	9	GgGGggGgG	FBKKKKKKK
20	36052986	N	9	AaAA$aaAaA	FFKKKKKKK
20	36052987	N	8	GgGggGgG	F/KKKKKK
20	36052988	N	8	GgGggGgG	FFKKKKKK
20	36052989	N	8	GgGggGgG	FFKKKKKK
20	36052990	N	8	AaAaaAaA	FFKKKKKK
20	36052991	N	8	GgGggGgG	FFKKKKKK
20	36052992	N	8	CcCccCcC	FFKKKKKK
20	36052993	N	8	TtTttTtT	FFKKKKKK
20	36052994	N	8	GgGggGgG	FFKKKKKK
20	36052995	N	8	AaAaaAaA	BFKKKKKK
20	36052996	N	8	GgG$ggGgG	BFKKKKKK

For site 36052981, if I provide the reference only two bases are reported, whereas without the reference 11 bases are found.
Note that quality bases also change.

I checked with IGV that the second case is the correct one (see snapshot). Interestingly, all but two of the reads that overlap 36052981 have an indel three bases before, so I guess the bug may be related to this. 

I am attaching a very small bam of the region. If it helps diagnosing the problem I can send more information for other “strange” cases.

Regards,
Federico