BWA aligned sam file missing "QNAME" format

[silentio]

Hi all,
I am using BWA to align pair-end read to hg19 genome.
Command line are as follows:

bwa aln $ref_dir/hg19 -t 4 $work_dir/${sample}_human_1.fastq.gz > $sai_dir/${sample}_1.sai
bwa aln $ref_dir/hg19 -t 4 $work_dir/${sample}_human_2.fastq.gz > $sai_dir/${sample}_2.sai
bwa sampe -r "@RG\tID:$sample\tLB:$sample\tSM:$sample\tPL:ILLUMINA" $ref_dir/hg19 $sai_dir/${sample}_1.sai $sai_dir/${sample}_2.sai $work_dir/${sample}_human_1.fastq.gz $work_dir/${sample}_human_2.fastq.gz | gzip > $sam_dir/$sample.sam.gz


However, some lines in the generated sam file are missing the "QNAME" field, starting from the FLAG field directly. As a result, the downstream command such as converting sam to bam cannot be done, giving information such as "Error parsing text SAM file. Not enough fields".

I wonder if it is the problem of bwa, or the problem of my fastq files?

Thanks in advance for your suggestions!

[swbarnes2]

Is gzipping the .sam really the right way to make a bam?

Try

bwa sampe all that stuff | samtools view -bSh - > sample.bam

[maubp]

However, some lines in the generated sam file are missing the "QNAME" field, starting from the FLAG field directly. As a result, the downstream command such as converting sam to bam cannot be done, giving information such as "Error parsing text SAM file. Not enough fields".

I wonder if it is the problem of bwa, or the problem of my fastq files?

Can you show us some of these bad SAM lines? Perhaps they have empty strings as the QNAME (which is bad), in which case can you look at the FASTQ file to see if that had records without an identifier?

[silentio]

I found the problem. It is because of the fastq files. Normally a fastq file starts with '"@" in the ID line. However, before running bwa, I used another software to do some preprocessing, which removed the "@". After adding "@" back, all return to normal.
Thanks for your reply anyway!