Subsampling from one paired-end fastq file

[morning latte]

Hi,

I know there were already few discussions about this topic but I am not sure I got it.

I have a fastq file containing Illumina paired-end reads. Below are the first four headers of the fastq file.

@HWUSI-EAS1599:82:64H78AAXX:7:100:10000:10533 1:N:0:CGATGT
@HWUSI-EAS1599:82:64H78AAXX:7:100:10000:10533 2:N:0:CGATGT
@HWUSI-EAS1599:82:64H78AAXX:7:100:10000:10642 1:N:0:CGATGT
@HWUSI-EAS1599:82:64H78AAXX:7:100:10000:10642 2:N:0:CGATGT

It looks like first and second lines are a pair and the third and fourth lines are the another pair. If I want to make a subset containing 1,000 reads, can I just extract the first 1,000 reads in order using 'head' command? I do not understand why it might cause biases. If 'head' command is not a good way for subsampling, any very simple way to do it? Thank you a lot for your comments in advance.

[dpryan]

Head isn't ideal since the reads near the beginning of the fastq file tend to be crappier. So, you really want to randomly subsample from the whole fastq file. You should be able to adapt the scripts found here and elsewhere to your case of having the pairs in the same file.

[swbarnes2]

The first tile in your fastq is going to come from the edge of the flow cell, so won't be as good as tiles in the middle. Use grep | head to get the first 1000 reads from a tile in the middle; that would be better. The tile ID should be the number after the lane.