How To Filter The Snps/Indel Results From Mpileup?

Question

[Haiping]

Hi. I trid to use mpileup to identified the SNP/Indels. The commands I used were:

samtools fillmd –bAr sample.sorted.bam ref.fa > sample.sorted.baq.bam

samtools mpileup -uf ref.fa aln.bam | bcftools view -bvcg - > var.raw.bcf

bcftools view var.raw.bcf | vcfutils.pl varFilter -D 100 > var.flt.vcf

I saw some of the SNPs that Qual value that low than 20 and Indels that low than 50.The problem is that I don't know if I can just trust this output resutls ar I still need to filter the resutls to make it reliable. If I need to filter, what kand of rule I should use? Thanks a lot.

Here are some of the results:

   1. 1102 . C T 3.01 . DP=10;AF1=0.4997;AC1=1;DP4=1,7,2,0;MQ=60;FQ=4.77;PV4=0.067,1,1,1 GT:PL:GQ 0/1:30,0,147:28
   2. 14689 . A G 18.1 . DP=10;AF1=0.5;AC1=1;DP4=0,6,0,4;MQ=60;FQ=21;PV4=1,1.1e-05,1,1 GT:PL:GQ 0/1:48,0,116:51
   3. 9373 . C A 44 . DP=10;AF1=0.5;AC1=1;DP4=0,6,4,0;MQ=60;FQ=47;PV4=0.0048,0.014,1,1 GT:PL:GQ 0/1:74,0,115:77
   4. 6427 . T TT 14.6 . INDEL;DP=9;AF1=0.5025;AC1=1;DP4=0,1,0,2;MQ=56;FQ=-14.7;PV4=1,1,0.12,1 GT:PL:GQ 0/1:52,0,20:23
   5. 314 . AA A 18.5 . INDEL;DP=9;AF1=0.5;AC1=1;DP4=4,0,4,0;MQ=60;FQ=18.5;PV4=1,0.0011,1,1 GT:PL:GQ 0/1:56,0,56:56
   6. 6068 . GATTAG G 214 . INDEL;DP=9;AF1=1;AC1=2;DP4=0,0,2,7;MQ=60;FQ=-61.5 GT:PL:GQ 1/1:255,27,0:51

Answer 1

[Swbarnes2]

Your depth of coverage is kind of low. Also, most of these SNPs displayed are mixed. So it's possible they are real, but I'd be skeptical. (But I don't have much empirical sanger to back that claim up) The last one, that has a decent quality score, because it has 9 reads that all agree that's an indel.

[Haiping]

Thanks for you response. The average depth of my data were near 30. Here are just part of the resutls. I just wonder that do I need to filger some of the results like that for 1,2,4 and 5 since the qual are less than 20 for SNP and 50 for Indels.

[Haiping]

Thanks for you response. The average depth of my data were near 30. Here are just part of the resutls. I just wonder that do I need to filter some of the results like that for 1,2,4 and 5 since the qual are less than 20 for SNP and 50 for Indels

Answer 2

[Travis]

The Broad Institute would probably recommend filtering based on recalibrated variant scores for a relatively low coverage experiment like this. Have a look here.

Answer 3

[Leszek]

Beside mentioned filtering, I often discard calls that are confirmed by alignments from one strand only as this is likely due to sequencing errors. Have a look at this discussion.