The GATK Best Practices without a dbSNP file?

[oiiio]

I have a BAM I am processing in GATK, and unfortunately I cannot provide a dbSNP for it.

Following the GATK Best Practices v3, I am trying to find a workaround for the absence of the dbSNP. Basically here is the plan,

1. indel realign
2. UnifiedGenotyper to produce a VCF list
3. Base Recalibration with the above VCF list for --known
4. UnifiedGenotyper again with the recalibrated BAM, and then analyzing the variants from this new VCF.

Does anyone have any criticisms/suggestions/comments about this procedure? I'm not entirely confident it is the best method.

[Heisman]

Is this a non-human genome and hence there is no dbsnp database? It's easy to download a dbsnp database otherwise.

Your strategy is fine if you only feed in the variants that you feel are real. I have no idea what good thresholds would be for real and not real, but if you are feeding in the real variants it will work. If you include a lot of non-real variants it will not only lose it's purpose but it will end up giving you more confidence that those false positives are real.