Hi,

I try to use samtools in a pipeline for variant calling from RNA-seq reads. To 
improve the quality of the calls, I preprocess the mapped reads with a tool [1] 
that splits reads containing a skipped region (N in cigar string) and removes 
parts of the alignment which are very short, e.g., from 93M1394N7M the 7M is 
removed by changing the alignment to 93M7H. When calling variants on that 
edited file, samtools mpileup misses a large number of SNPs which should at 
least be reported. I narrowed it down to the calculation of the BAQ so I read 
your papers about it but still do not understand why the BAQ for the bases is 
lowered so much. Please find attached SAM files showing the original reads and 
the “splitted reads”. I added a visualization from the IGV in which I marked 
the questionable SNV by a red arrow (pos chr12:85277730). Note that I have 
chosen this example because of the low number of reads and that examples with a 
higher coverage exist.

The pileups using different parameters display a low BAQ for the mutation after 
the splitting but not in the original files:
$ samtools mpileup         -f genome.fa -r chr12:85277730-85277730 original.bam
chr12     85277730             C             15           ,......,...,TTT       
  FHEJJIIDJHGCDFF

$ samtools mpileup         -f genome.fa -r chr12:85277730-85277730 splitted.bam
chr12     85277730             C             12           ,......,...,          
      FHEJJIIDJHGC

$ samtools mpileup    -Q 0 -f genome.fa -r chr12:85277730-85277730 splitted.bam
chr12     85277730             C             15           ,......,...,TTT       
  FHEJJIIDJHGC!!!

$ samtools mpileup -B      -f genome.fa -r chr12:85277730-85277730 splitted.bam
chr12     85277730             C             15           ,......,...,TTT       
  FHEJJIIDJHGCDFF

The behavior is reproducible with samtools 0.1.19 and 1.2

To conclude, I do not understand this severe drop of quality just because the 
last 7 bases of the reads are trimmed. Can you probably explain this or do you 
recommend to not using the BAQ?

Thanks for your help in advance
Manuel

[1] https://github.com/mjafin/splitNreads (results equal GATKs SplitNCigarReads)



[cid:[email protected]]

Dr. Manuel Landesfeind
Bioinformatician
Oncotest GmbH, Am Flughafen 12-14, 79108 Freiburg, Germany
phone: +49 (0) 761 51559 591 ▪ fax: +49 (0) 761 51559 55
email: [email protected]<mailto:[email protected]>
website: www.oncotest.com<http://www.oncotest.com/>


This message and any attachments are confidential and may be privileged or 
otherwise protected from disclosure. lf you are not the intended recipient, you 
must not copy this message or attachment or disclose the contents to any other 
person. lf you have received this transmission in error, please notify the 
sender immediately and delete the message and any attachment from your system.



Attachment: original.sam
Description: original.sam

Attachment: splitted.sam
Description: splitted.sam

------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that
you need to offload your IT needs and focus on growing your business.
Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
_______________________________________________
Samtools-help mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/samtools-help

Reply via email to