How to do unique mapping in tophat2 for paired end?

[metheuse]

I know there is a -g parameter to control how many hits to report. But I want to discard the reads that could be mapped to multiple locations. Bowtie2 has a parameter -m to control this. But I can't find a similar thing in tophat.
Any help?

[dpryan]

Just filter after alignment. Tophat gives reads with unique alignments a MAPQ of 255.

[metheuse]

Just filter after alignment. Tophat gives reads with unique alignments a MAPQ of 255.

Thanks.
Yes, that's what I'm doing now. Actually the mapq for unique mapping is 50 now, in tophat2.

But there is something I don't understand:
I ran tophat2 with different -g (1 and 20) and selected the hits with mapq=50, the numbers differ. The number from -g=1 is fewer than from -g=20, and the former is not a subset of the latter. I don't understand why they are different, and I don't know which is right.

[Aharon]

What a nice coincidence.

I am doing these days my first attempt to analyze paired end sequencing data (not that i did single end before:-() and just understood that i am getting very bad results because of reads which have multiple alignments.

So in filter after the alignment, you mean to get into the scripts of the Tophat, and do this after the Bowtie2, before the next steps are taking place? I guess is will be much less efficient to do it on the end result.

How come that there is no such a flag in the Tophat? Isn't it more or less a must to do such a filter while doing paired end?

[Aharon]

I cannot find an analogue to the -m 1 option of Bowtie1 in Bowtie2. Any idea?