I am doing a project that requires bench-marking the performance of computational tools for alignment and variant calling of human genome sequencing data. In particular I would like to establish benchmarks comparing the performance of GATK and ADAM/Avocado.

My main question is which genomic data-sets should I use as the input to the data processing pipeline. In Frank Nothaft's (author of ADAM and Avodaco) thesis he reports bench-marking the processing of NA12878 from the 1,000 genome's project in Avocado as compared to GATK's HalotypeCaller. He also reports benchmarking on 270 samples from the Simons Genome Diversity Data-set.

I am interested in reproducing this portion of his work, but do not know where I can get a hold of these datasets. Where can I get a copy of the NA12878 genome (BAM file) and also were can I get aligned samples from Simons Genome Diversity Data-set?


1 Answer 1



  • My recommendation: download raw reads from Illumina BaseSpace and do alignment yourself. Aligning ~40X worth of human reads takes overnight with ~16 CPU cores.

  • You can acquire Platinum Genome data from ENA. There is a BAM available for download, but that was done by Illumina's own aligner against a wrong reference genome. You'd better do alignment by yourself. Another problem with this dataset is it is too good to be representative. Modern data are worse.

  • GIAB provides BAMs here:


    It is not "typical", either.


  • Instructions on downloading SGDP BAMs are available on this page, in particular this Word document. Warning: you need ~30TB disk space for these BAMs. I also heard it is tricky to download data.

  • There is apparently a mirror from cancer genomics cloud. I don't know how easy it is to access.

  • $\begingroup$ I'm having some trouble finding the fastq files for NA12878 on Illumina BaseSpace. Do you have any suggestions on where to find these? $\endgroup$
    – Jon Deaton
    Commented Jun 25, 2018 at 17:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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