Considering the cost and the achievement of the methylation sequencing, Do you think single-end RRBS or BS-seq would be better than pair-end RRBS or BS?
Here is my reason:
There would be overlap between pair-end sequencing, therefore, loss would be happened here.
Pair-end alignment process would bring large number useless mapping/reads
But maybe pair-end RRBS have higher creditable result since the alignment accuracy would higher than single-end RRBS.
Am I right? Any comments? Is there any strategy to quantitatively assess this question?
This will depend on your read length. If you start doing RRBS with 2x250 reads or something like that then you're correct that read #2 is largely wasted. I've personally mostly dealt with 2x50/2x36 (my older RRBS datasets) and 2x100 datasets (WGBS), where the overlap is pretty minimal. If you're doing a lot of RRBS then you might look at the fragment distribution that you're enriching for and determine the right read length and whether to spend the money and time on read #2 accordingly.
BTW, this assumes that you're still getting good quality base calls near the end of the longer reads. The benefit to having two reads covering the same position is that you get a more reliable methylation call (this is the case with PileOMeth, at least).
Edit: A strategy to look at this might be to "samtools depth" the files with and without setting a phred score threshold. samtools depth should change the phred scores to 0 for one of the mates where there's an overlap (and the overlaps agree), so you can take advantage of that.
