tophat with mixed paired end reads

[nimmi]

Hi,

I have a file with mixed paired end reads (50bp and 70bp). My insert size is 400bp. Did anyone try running tophat with mixed paired end reads? If so, what would be the tophat -r option (will it be 300bp or 260bp or an average 280bp)?

Thank you in advance.

Nirmala

[GKM]

In general, I don't think mixing different read lengths is recommended, although I see that 1.1.2 is supposed to support variable length reads, but I haven't tried it and I am not exactly sure what this means, i.e. does it mean you can mix the 27bp with the 100bp as the manual explicitly advised against before (and still advises against), or it only works well with a more limited range of variation.

How many reads do you have? I would try running it on mixed reads, with an -r of 280, and if the output is anomalous (Cufflinks can't assemble anything, you're missing a lot of junctions, etc.), switch to something like this: you could trim all reads to 50, map, get junctions, map all 70s separately, get the junctions again, add the annotation, and them map the 50s and 70s separately against the resulting set of junctions, and then join the output.

[zach]

Tophat output bed file

I have some naive question in terms of the output from TopHat.

track name=junctions description="TopHat junctions"
chr start end name score strand thickst thickend xxx blockcount blocksize
test_chromosome 180 402 JUNC00000001 46 + 180 402 255,0,0 2 70,52 0,170
test_chromosome 349 550 JUNC00000002 38 + 349 550 255,0,0 2 51,50 0,151



Question: 1) what kind of score (here 46 and 38) is a good indicative of splicing junction
2) the block count is the read counts on the two sides of a junction? e.g. 70, 52 means 70 reads on the left and 52 reads on the right of the junction
3) what is the first number 0 in the block size 0,170. In this case I assume that the block size is 170 nt. And therefore the gap (intron) between this junction is 52 based on the calculation 402 - 180 = 222 and then 222 -170 = 52.