Cram

[ulz_peter]

Hi Guys,

Anyone already thought about deploying CRAM for archiving alignment data, rather than BAM?

http://www.ebi.ac.uk/ena/about/cram_toolkit

Our Illumina representative told us they want to switch to that format...

[maubp]

I've heard Sanger is considering it, perhaps even this year if CRAM continues to mature rapidly.

[lh3]

Yes, cram has a great potential. It may ultimately replace BAM (if cram does not do that, there will be a binary format to achieve sooner or later). Nonetheless, cram may not replace BAM right now. It does not (at least did not) support all the tags. I do not know the progress on compressing unmapped reads. Furthermore, I am concerned with the compression model. I also think lossy compression is the way to go, but this should be done by reducing the resolution of quality, instead of by selectively dropping all the quality information.

[maubp]

Yes, cram has a great potential. It may ultimately replace BAM (if cram does not do that, there will be a binary format to achieve sooner or later). Nonetheless, cram may not replace BAM right now. It does not (at least did not) support all the tags.

Supporting all the tags is expected in CRAM 0.7 due soon, see e.g.

http://lists.open-bio.org/pipermail/bioperl-l/2012-March/036295.html

I do not know the progress on compressing unmapped reads.

I heard at a recent seminar that the CRAM team are looking at doing a mini-assembly of the unmapped reads in order to generate dummy reference sequences which can then be used for reference based compression. If I understood correctly this might be transparent to the user.

Furthermore, I am concerned with the compression model. I also think lossy compression is the way to go, but this should be done by reducing the resolution of quality, instead of by selectively dropping all the quality information.

Also at the same seminar we were told CRAM has several modes of quality compression, one of which is simply reducing the resolution.

[lh3]

That is great! Hope these can be done soon!

[cjfields]

That does seem very promising.

[vadim]

I am the developer of CRAM and can answer any questions about it.

The code is here:

https://github.com/vadimzalunin/crammer/

Documentation can be found here:

http://www.ebi.ac.uk/ena/about/cram_toolkit

We just released v0.7, which is not a long term support yet but stable enough to try it out.