Samtools mpileup generate different SNP result every run?

[edge]

Hi,

I just wondering is it normal that re-run exactly the same data set by using exactly same command will generate different SNP calling result?

Currently I notice that if I re-run SNP calling by using samtools mpileup.
It gives me different result every run
As my data input and command used is exactly the same all the time.

Thanks for any advice.

[rfilbert]

I have noticed this too. Pretty disconcerting...

[edge]

Hi,

Do you get any feedback or advice from Samtools to solve this issue?
It really surprised me when I re-run exactly the same data set and same command for few times.
Every time it gives different result

[rfilbert]

Does Samtools offer technical support?

[edge]

I have no idea about it too

I just reported it as a bug of samtools mpileup.
I test the samtools mpileup for genome and transcriptome data set, both still same

Just wondering GATK got same issue too?

[xied75]

I did a search of the source code, there is a srand48(time(0)) over there, haven't got time to trace the code but this could be the reason because it was using current system time as random seed.

[rfilbert]

i guess sometimes we get what we pay for...

[rfilbert]

In the samtools source there is only one instance of calling srand48 with the current time as seed (which will result in different results each time you run it).
It is only called if you run “bcftools view” with the -U parameter greater than 0
I guess you have to be a programmer to use these tools...

[kmcarr]

I just wondering is it normal that re-run exactly the same data set by using exactly same command will generate different SNP calling result?

Out of curiosity how different is different? Are we talking huge numbers? Are the differences between SNPs called with an extremely high score in one instance and not called in the other or are the differences all for SNPs in the "twighlight" of reliability? Are the differences before or after filtering the raw SNP calls?