Do I need to change tophat parameter if reads length distribution is odd?

[super0925]

Hi all, my sample's reads length distribution is odd (there are too many short reads) I uploads two figures , one is sharp ( Q1: I think it is good quality am I right? ), one is not so good (lots of short reads compared with longer reads).

Q2: Do I need to change tophat parameters when doing the mapping? or just use the default setting?

cause from biostar I know , tophat first try to align the whole reads. After that it cut the unmapped reads in little segment (at least 25 bp length by default) and try to map it again.
https://www.biostars.org/p/52033/

I don't want to trim the reads too much cause it would impact my final result, am I right?

Thank you!

[Brian Bushnell]

What do these graphs mean, and how did you generate them? What platform are you using? With Illumina, for example, all reads come out the same length. Are these Ion Torrent?

[super0925]

What do these graphs mean, and how did you generate them? What platform are you using? With Illumina, for example, all reads come out the same length. Are these Ion Torrent?

Yes.
you are right!
My question, do I need do more trimming more on this sample? (cause default 'genetle trim' in RNA-Seq, so I only remove the adaptors.)
And do I need change parameter in Tophat?
Thank you!!

[Brian Bushnell]

The best way to determine if you need more trimming is by looking at e.g. FastQC plots to see if there are overexpressed sequences, or to map and look at the error-rate on the 3' end of the read (or wherever you expect the adapters to be).

With BBMap, you can do this:

bbmap.sh ref=reference.fa in=reads.fq out=mapped.sam mhist=mhist.txt qhist=qhist.txt nodisk

Then plot the mhist.txt data in something like Excel, and it will show you the error rate per base position. If the error rate is higher than expected or allowable on the 3' end then more trimming is probably useful.

[super0925]

The best way to determine if you need more trimming is by looking at e.g. FastQC plots to see if there are overexpressed sequences, or to map and look at the error-rate on the 3' end of the read (or wherever you expect the adapters to be).

With BBMap, you can do this:

bbmap.sh ref=reference.fa in=reads.fq out=mapped.sam mhist=mhist.txt qhist=qhist.txt nodisk

Then plot the mhist.txt data in something like Excel, and it will show you the error rate per base position. If the error rate is higher than expected or allowable on the 3' end then more trimming is probably useful.

Hi
I have attached my FastQC report. one is good (>=20), this is ion-torrent but not Illumina data, so >=20 is OK, the other is not good. Currently I just remove the adaptors and trim the length >=16.

Q1: What do you think , is the bad quality sample need trim?
Q2: If trim,what length do you prefer?
Q3: Do I need to reset the paremter in Tophat?

Thank you!
Could I change any parameters in trimming

[Brian Bushnell]

I am hesitant to give you any suggestions on how to trim Ion Torrent data, with which I have no direct experience; hopefully someone who has dealt with it will chime in.

[super0925]

I am hesitant to give you any suggestions on how to trim Ion Torrent data, with which I have no direct experience; hopefully someone who has dealt with it will chime in.

How about Illumina? Thank you.
I just want to ask (1) do I only need to do more trimming if reads quality is not good or (2) if I change Tophat parameter (like segment length) , is that useful or impact the result?