TopHat: cutting reads into segments

[nike00]

Dear all,

as you know TopHat cuts each read into multiple segments whose size is 25b by default.
From the manual, this is the corresponding option:
--segment-length #Each read is cut up into segments, each at least this long.. The default is 25.

Since I'm working with (pair-end) reads of 70b, I can imagine the reads cut into 3 segments: 25, 25, 20.
Is it right? or the shortest (20) is not considered because of "each at least this long."?

So, in the end, is the default 25 a good choice? or it's better to set a segment-length shorter, for example, 23, to obtain: 23,23,24?

Thanks in advance for any suggestion,
nike00

[chadn737]

The tophat creators only recommend decreasing segment length if the reads are ~45 bp or under.

I know from personal experience, I was working with a sample of mixed lengths (32-100bp) because I had trimmed based on quality. I tried decreasing the segment length because of the 32 bp reads, but actually got worse results than when I ran it on default.

[nike00]

thanks chadn737 for your reply,

I would like to know more about it.
Anybody knows how TopHat manages the segments?
is each of them considered, in any case? or are they filtered out depending on their shortness?



thanks