Hi Broad folks, I am looking for a little bit of guidance recalling some bams (which were actually sequenced at the Broad as well). I have been working through the workflow described here, which has been very useful. I have two questions I am looking for some help on:
Question 1. While working through this, I discovered that the SM tag was different for different reads in the same sample. For example:
Read1:
H7FL7BBXX160404:2:2115:11809:26494 99 1 10004 18 3S60M1I9M3S = 10359 429 "sequence redacted" @>?>?@B@@B@@CAAB@@CAAB@@CAAB@@CAAB@@CAAB@@AAAB@@CAA?@@@AAB@@:AAB@@@;A>A>==@= MC:Z:2S74M MD:Z:69 PG:Z:MarkDuplicates.A RG:Z:H7FL7.2AM:i:18 NM:i:1 SM:i:18 MQ:i:18 OQ:Z:AAFFFJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJFJJJJJJJJFJJFJJJJJ7JJJJJJ<JFJJJJJF UQ:i:0
Read2:
H7FL7BBXX160404:2:1124:24282:45361 163 1 10005 15 27M1I48M = 10247 318 "sequence redacted" <:>>?@>>A@@=>>A@@@>>A@@@>>>@AA???B@@A??>AAA@@CAAB@?CAAB@@@AAB@@C1AA??9?>><=# MC:Z:76M MD:Z:3A5A5A5A6A10A35 PG:Z:MarkDuplicates.ARG:Z:H7FL7.2 AM:i:15 NM:i:7 SM:i:15 MQ:i:15 OQ:Z:AAFFFJJJJJJFJJJJJJJJJJJJJJFJJJJJJJJJJJJAJJJJJJJJJJJJJJJJJFJJJJJJ<JJJJ7JJJJJ# UQ:i:242
In the reads above, I do not understand why the first read has SM:i:18 while the the second has SM:i:15. Shouldn't these be the same? For tools we're using downstream, they require that we give one SM tag for each sample (to, for example call variants between them) - is it possible to just reassign every SM tag to say, for example "Tumor" or "Normal" - or would that screw with BQSR or some other downstream step? (We are using GATK best practice for the entire pipeline). In general, the read group information appears to be independent of the SM information, so this is not duplicate information.
Question 2: While the sequencing was done at the Broad institute, different capture kits were used - around 50% of the samples use the Agilent SureSelect v2.0 while around 50% use Illumina's Rapid Capture Exome Kit (version unknown - it says 38Mb targeted). Do you have any standard advice for how to unify the data from these two different kits? Should we be only using those reads which map to regions covered by both? Or should we keep the data from different kits separate?