Difference between "bedtools bamtofastq" and Picard's SamToFastq

[thedamian]

Hi,
I need to assembly some unmapped reads from Tophat2's output "unmapped.bam"
Firstly I have to extract read pairs from unmapped.bam

I found two tools for it:

bedtools bamtofastq -i unmapped.bam -fq bedtools_un1.fastq -fq2 bedtools_un2.fastq

and

java -jar SamToFastq.jar I=unmapped.bam F=picard_un1.fastq F2=picard_un2.fastq VALIDATION_STRINGENCY=SILENT

results:
bedtools_un1.fastq has 685 MB
picard_un1.fastq has 915 MB

My question is, why the difference in the size of files is so big?
Shouldn't the result files be the same? (as the inmput "unmapped.bam" is the same)

[TiborNagy]

Please take a look inside the files. It should be the same size, but maybe one of the programs write out different header or give full header in the quality section.

[thedamian]

wc -l picard_un1.fastq
15367376 picard_un1.fastq

wc -l bedtools_un1.fastq
11515208 bedtools_un1.fastq

headers are the same

[dpryan]

Tophat will often map only one mate from a pair. In those cases, it's up to the program to determine whether such reads should or should not be written to a file. It's quite likely that SamToFastq outputs orphaned reads whereas bamtofastq doesn't.