genomicsaotearoa / dbnsfp_build Goto Github PK

The fix to issue #1 has revealed another 'bug', this is to do with positions differing between genomic builds.

Example: after sorting chromosome 6 based on hg19 positions their are 639 variants that are annotated to chromosome Y (meaning these 639 variants are annotated as chromosome 6 under hg38).

Alongside chr 6 this has been noted on chrs 17 (has chr 11 variants) and 22 (has chr 14 variants).

This issue was discovered because the implemented fix to #1 involves processing dbNSFP data at the individual chromosome level rather than trying to perform a sort on one large compressed file (as was done previously). This means in it's current implementation the combined output of the build script ends up unsorted and unable to be indexed using tabix.

• - develop testing dataset using three smallest chromosomes (13, 19, 21) so that sorting across chromosomes is tested
• - write test which compares output to a known-good output
• - implement TravisCI for testing commits

@sirselim has found a bug in the hg19 workflow that he plans to fix.

Hey @mikblack where could we host this (the 25GB database output file) in NZ (on some kind of REANNZ linked location) so that it could be readily available to researcehrs. @sirselim pointed out to me that tabix and its associated index of ~850k can be used to query the parts of the output database from an https:// location. This may be a good demonstration project to show how GA can roll out infrastructure which is useful to other groups. We can track usage of these things as a metric you can use for reporting back to funders as well.