Reads mapped on junctions

[nike00]

Hello everybody,


I work with Tophat to align paired-end reads.

By reading one of the log-files that it produces: reports.log,
I can know the number of the Loaded junctions and the Found junctions.
Does anyone know how can I count the reads mapping on the Found Junctions?


Thanks,
nike00

[cedance]

There should be a file "junctions.bed" that gives you a detailed information regarding the junction coordinates and the number of reads.

Alternatively you could do it yourself from the SAM/BAM file, if you go through the SAM format manual. If you would like to do it by yourself, if necessary, I'll explain in detail.

[nike00]

Dear cedance,

thanks a lot for your reply and suggestions.
I have checked the file "junctions.bed" and, as well as I have understood, there is a column (the fifth one) reporting a score. The score is the number of alignments spanning the junction (from the Tophat Manual).
It exactly is the number of pairs (of reads) spanning the junctions, isn't it? or the total number of the reads?

PS: For sure I'm going to explore the SAM file by myself, so I will post again if necessary.

Thanks very much,
nike00

[cedance]

nike00, I am glad I could be of help.
I think its the total number of reads that span the junction. The other pair of the read (that is mapped across this junction) need not span the same junction, as it depends on the fragment length (or inner distance as Tophat says it) you chose (after adapter + optionally barcode ligation).

Therefore, the number given by junctions.bed is the number of reads spanning the junction and not necessarily its pairs.

Paired end reads, imho, are used for better mapping efficiency. You could find out the number of reads where both pairs map to the same junction, but I don't see why or what is the use.

I'm in bit of a hurry, if what I wrote is more confusing, please reply back and I'll try to explain better what I meant/think.