Align paired and unpaired reads with Tophat
 

I have a stranded PE RNAseq data set that I want to align with tophat using the --library-type fr-firststrand option. After adaptor trimming I end up with my 4 files:

paired_1
paired_2
unpaired_1
unpaired_2

Is there a way to align these so I do not loose the strand information for the unpaired reads? When I run tophat with the files listed like this:

paired_1,unpaired_1 paired_2,unpaired_2

It seems to want to try and align the two unpaired files as paired files. If I combine the two unpaired files and run tophat with the files listed like this:

paired_1 paired_2,unpaired

It recognizes the last file as unpaired. But am I loosing my strand specificity by aligning this way? Thanks.

Inputting the files like:
Will result in "unpaired_1" and "unpaired_2" being treated as paired, which is the opposite of what you want. To keep the strandedness correct you'll need to run things twice. Firstly using "paired_1,unpaired_1 paired_2" with library-type set to fr-firststrand and then just "unpaired_2" by itself with "fr-secondstrand". I should note that aligning unpaired_2 is usually not worthwhile (the reads are often crap), but perhaps you'll get luckier than I have with that.

Thank you very much for your response, that makes a lot of sense.

On this note ..I Have a question. I am doing PE RNA-Seq analysis of mouse data. My read length is 90bp and fragment size is 127 bp which means I have overlapping reads. I used Flash and found that not all reads overlap.

I have a file with merged reads which overlaps and also two files with non overlap reads. basically three fastq files.

How do I go about running Tophat with this and I am not really sure hot to calculate the mean inner mate distance and sd??

Has anyone come across this situation???

There's no need to run Flash on them, just use tophat2 and bowtie2 instead of bowtie1.

For the mean inner distance, just try 0 and see if that produces acceptable results (I recall reading that tophat re-estimates the insert length as it runs, though I can't say I've ever checked if that's correct).

Thanks alot Ryan... I will try that.

or just use STAR that do not need to specify inner distance. and also is much faster for the same ( even better ) results