TopHat local alignment

[Giffredo]

Hello!

I know that TopHat program run the alignment Bowtie end-to-end like first alignment.

My question is: is it possible set local Bowtie alignment in Tophat?

[kmcarr]

From the TopHat manual

Preset options in --end-to-end mode (local alignment is not used in TopHat2):

so it appears the answer to your question is no.

[Giffredo]

Thanks for the reply.

yes I read the manual, I hoped in same tool or something able to give me a result similar to local alignment...
My problem concerns the start of my reads has a low quality and the trimming cannot help me because of the tophat limit to map read >75base
If I trimmed the reads of 10 bases (number suggested from fastQC tool) I would waste too many reads.

So far I reached a maximum of alignment of 20%

[dpryan]

Tophat isn't limited to reads >75 bases. I think the minimum is around 20 bases.

[Giffredo]

Are you sure?

i read this: "The software is optimized for reads 75bp or longer."

[westerman]

As far as I know, Tophat does alignments in 'segment' sizes of (by default) 25 bases and then stitches the alignments back together in order to figure out if a read maps to the genome (and possibly crosses an intron/exon boundary) . Since Tophat is working with short length segments then it only makes sense to do end-to-end mapping and not try local mapping.

If you have poor quality sequences then modifying some of Tophat's parameters should allow more mapping. I suggest:

--read-mismatches ... default is 2, try something like 6
--read-edit-dist ... default is 2, try 6
--read-gap-length ... default is 2, you may wish to go higher on this

--segment-mismatches ... default is 2 (out of 25), going to 3 may be useful
--b2-N ... default is 0; go to 1 to allow more mismatches

[westerman]

Are you sure?

i read this: "The software is optimized for reads 75bp or longer."

Optimized is not a hard limit.

That said I am not sure what the lower limit is. I suspect the segment size (default 25).

[dpryan]

Yup, it's optimized for longer reads, but that's not the minimum read length. Shorter reads will just be less likely to align, particularly if you're looking for novel splice sites (since the read has to be split and the portions aligned individually). This is also related to Rick's post right above mine about why it doesn't bother with local alignment (if you really want that, just use STAR, which is faster anyway).

Edit: @Rick, yeah, 25 sounds right.

[Giffredo]

Thank Rick and Ryan very much!

I read about STAR but I have to stay with Tophat (Anyway by now it is a "friend" for me )

the function of --read-edit-dist parameter is not actually clear..... ....