Remote FASTA reference support?

[kyleabeauchamp]

I noticed that samtools view -T http://genome.fa.gz my.cram fails with an exception:

[samfaipath] fail to read file http://genome.fa.gz.
http://genome.fa.gz: No such file or directory

I'm curious what the "optimal" workflow is for handling remote reference genomes with operating on CRAMs. Obviously I can do a manual copy of the reference first, but I'm just curious if that's the recommended workflow.

[daviesrob]

No that doesn't work (even with a valid URL). The code that loads cram references from fasta files makes calls to functions like stat() and fopen() which only work on local files. There are plans to replace this with a version that uses hfile throughout which would allow loading over http.

It is possible to get references using the MD5 checksum via http if you have @SQ headers with M5 tags (cram files should have these). The default is to try to get them from the EBI server and then cache them locally (to reduce load on the EBI), but if you set REF_PATH to point to your own server that takes the MD5 and serves up the corresponding sequence it will work.

[kyleabeauchamp]

SGTM, I'll close this.