Entering edit mode
22 months ago
optimistsso4co3
▴
130
Pipeline i analyse uses three steps for recalibration:
- BaseRecalibrator (in parallel)
- GatherBQSRReports
- ApplyBQSR
Is it appropriate to skip 2nd step and applyBQSR on each chunk separately with separate recal.table's ?
one you have computed BaseRecalibrator , GatherBQSRReports takes only a few seconds, so there is no use in skipping that part IMHO.
Beyond that, the reports are effectively count tables; and gathering them adds the counts together. If you compute the recalibration table and apply it in the same chunk (say 1MB) then your quality scores will be horribly under-estimated, since the total number of counts will be insufficient for a score of >Q25 or so.