I am bench-marking the performance of ADAM/Avocado on NGS whole-genome sequencing datasets. I am trying to repeat the tests reported in Frank Nofthat's (author of ADAM and Avocado) thesis, however, I am running into an issue acquiring the dbSNP database to use for Base Quality Score Recalibration.

The Base Quality Score Recalibration command (page 128-129):

adam-submit -- transformAlignments aln.adam bqsr.adam -recalibrate_base_qualities -known_snps dbsnp.adam

requires dbsnp.adam, which is described as "the dnSNP database" (converted to ADAM format) with reference to the NCBI database of genetic variation. This "database" can be found at this FTP site however I do not know which parts of these files to download and convert to ADAM in order to perform this step of the transformation pipeline.

Also, in order to use this genetic variation databaes do I need to make sure that the reference genome of "known variants" is the same genome that the sequencing data was aligned to?

Thank you.


In the dbSNP FTP site you linked, you need to go into organisms and select your organism of interest (human obviously). You can then select which release of the SNP database you want (eg. 150 vs 151) [the modified dates hint at when they were released] and which genome build you want (eg. hg37 vs hg38).

Below I am looking at the .vcf files. I assume this what you want because they can be converted to ADAM format via the adam vcf2adam command.

For example human (newest genome and newest dbSNP release): ftp://ftp.ncbi.nlm.nih.gov/snp/organisms/human_9606_b151_GRCh38p7/VCF/

Your next decision is whether or not you want to use ALL the annotated SNPs in the database or just the common ones. For the former you choose the All_20180418 file and for the later the common_all_20180418 file. I'm pretty sure the convention here is that the date in the filename is when it was uploaded/updated.

In your case I'm not sure what makes the most sense. Firstly, are you trying to replicate the initial conditions of when that software was published or do you want simply to benchmark the method with the best available data. Besides that, I imagine using all them is computationally much more taxing and if you are just coming up with general trend then using the common ones seems viable.

  • $\begingroup$ This is super helpful. One more question though: does my input BAM file need to have been aligned to the same reference genome (in this case GRCh38)? If so I will need to re-align my input BAM to the CRCh38p reference genome before performing this step. $\endgroup$
    – Jon Deaton
    Jun 29 '18 at 21:33
  • $\begingroup$ Yes ideally although it looks like they also have an hg37/hg19 version $\endgroup$
    – story
    Jul 10 '18 at 11:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.