duplicate read names in TCGA raw bam file

Question

[jonessara770]

Hi,

I downloaded some TCGA ExomeSeq bam file and would like to convert them to fatsq. Picard can not convert them because there are some reads with identical same readname in the bam files. As suggested by the community bedtools bamtofastq can randomly choose only one of these reads and put it to fastq file. I tried bedtools bamtofastq but the duplicate read names are still in the fastq output.

My question is :

• Why there are reads with identical readname in bam files?
• How can I fix this issue? should I rename the reads name in fastq files?

Thanks Sara

[GenoMax]

Possible that the R1/R2 identifiers were stripped from the reads? Multi-mappers?

[jonessara770]

No there are there. this is output of bedtools bamtofastq

fastq1

@C0UL8ACXX120727:8:2301:17486:13986/1

@C0UL8ACXX120727:8:2301:17486:13986/1

fastq2

@C0UL8ACXX120727:8:2301:17486:13986/2

@C0UL8ACXX120727:8:2301:17486:13986/2

[GenoMax]

An identical issue was discussed in a recent thread. Edit: Unfortunately there was no clear solution (check out the thread yourself: Remove duplicates reads Ids ).

[mastal511]

I've had the opposite problem (bedtools bamtofastq did not give the required results, Picard tools did the job), but with single end data, so I don't know what would happen for paired-end reads.

The bam files had multiple entries for some of the reads, which were multi-mappers. With bedtools bamtofastq the multi-mapping reads ended up present multiple times in the fastq file. Picard tools ran without any errors and gave only 1 copy of each read in the fastq file.