Reads of different length?

[nike00]

Hi everybody,

I have a question for you.
I think it is very basic, but I have not a clear idea about it.

I'm working with paired-end reads (2x71) for RNA-seq analyses.
After sequencing, I converted the bcl files into fastq and I checked the Per Base Sequence Quality of each line and each mate of the pairs.

Now, I might trim the last bases of the reads without a good quality in each line, but I would get reads of different length for each line (depending on the quality) and I might do (theoretically) the same for each mate of the pairs.

At the end I could have:
1. reads of different length in different lines
2. reads of different length in the same pair

Is this kind of data good for the analysis?
which one should I avoid? both?

Thanks for any suggestion!!
nike00

[cedance]

Now, I might trim the last bases of the reads without a good quality in each line, but I would get reads of different length for each line (depending on the quality) and I might do (theoretically) the same for each mate of the pairs.

Yes, that's right.

At the end I could have:
1. reads of different length in different lines
2. reads of different length in the same pair

Yes, right again.

Is this kind of data good for the analysis?
which one should I avoid? both?

Yes, from my experience and logically, I find both of these imperative. However, I put a threshold on the Quality-trimmed reads. If they are >= 50, I retain it (I do it separately on each pair separately). Of course there are situations where one pair has <50 and other has >= 50.
I use tophat. I remove those reads < 50, and put the >=50 read in a separate SE file and do another tophat run with just the SE reads by feeding the junctions file from the previous run from PE reads.

Hope this helps.