indexcov: different coverage scale with cram and bam inputs

[kpalin]

With input bam files, the scaled coverage in indexcov is about 1.0 but with cram input the coverage is about 50000.

[brentp]

is this for whole genome? can you share a single crai file that has this problem?

[kpalin]

It's more complicated than I thought. Sent you two crai files (via funet filesender, you should have the email), one good, one bad.

[brentp]

I see what you mean about scaling.

goleft indexcov -d /tmp/tt --fai /data/human/g1k_v37_decoy.fa.fai ~/Downloads/bad.cram.crai

on the bad.cram.crai that you sent. I will look into this. For #37, could you send a bam for that as well?

[brentp]

I have found a fix for this. There was a problem with the calculation of median depth for scaling to 1 that only manifested in extreme cases like yours.

Would you try the attached binary and confirm that this resolves it for you?

thanks for reporting.

goleft.gz

[kpalin]

Now all crams work but almost all bams don't. Only 2 out of 119 have median value > 10^-5.

[brentp]

I fixed a bug that I introduced during that change and neglected to update the binary here. would you mind trying this one?
goleft.gz

[brentp]

please re-open if the problem persists with latest version.