Illumina paired-end reads strand detection

[imang]

Hi,

I'm working on Illumina short paired-end reads of human genome and when I do BLAT on short reads, I'm seeing that not all read /1's are from one strand. e.g. here is what I got for my first 5 paired-reads:
Read# R1 R2
(1) - +
(2) + -
(3) + -
(4) + -
(5) - +
...
So, my question is: regardless of reads orientation, is there any rule that first read is always coming from one strand and Read 2 from the other one in Illumina sequencing? I was assuming this but it seems that its not true.

I appreciate if anybody can comment on this.

[GenoMax]

Basically Yes. What you are looking at is mapping relative to a reference so it could be with either strand on top (unless directional libraries were made). See this: http://www.homolog.us/blogs/blog/201...looking-reads/

[imang]

Thanks for the link. So here is the thing I got: Even though we know the directionality of reads (inward/outward), we cannot sort the whole set of reads to be in only one strand (+ or -).

By doing the pre-computation described in the link, all paired reads become in one direction but again some pairs belongs to + and some to - strand. is this correct?

[Heisman]

That is correct. To separate the strands you can align the reads and then from the alignment files it is possible to pull out reads that align to one strand vs. the other.