Why there is a difference in size between R1 and R2 fastq files from BAM?

[genomics]

Hi,

I downloaded bunch of bam files from TCGA projects and converted them to fastq files. Even though these BAM files are pair-end WES data, the size of fastq_R1 is different with the size of fastq_R2 in many cases. Sometime, the difference are 1 or several lines as well as almost two fold differences in file size. I used SamToFastq.jar of Picard. The command I used is as follows;

java -Xmx16g -jar SamToFastq.jar TMP_DIR=. INPUT=$file FASTQ=fastqR2 SECOND_END_FASTQ=fastqR2 VALIDATION_STRINGENCY=LENIENT VERBOSITY=ERROR";

For example,
TCGA-50-5946-01A_R1.fastq -> 3504190138
TCGA-50-5946-01A_R2.fastq -> 3504224169

TCGA-A7-A0CE-11A_R1.fastq -> 40393996094
TCGA-A7-A0CE-11A_R2.fastq -> 21008959121

It would be great if you give a clue about this.

[Brian Bushnell]

R2 is generally lower quality and more likely to be trimmed if quality-trimming was done, and less likely to be mapped (if for example the bam only contains mapped reads).