Tweet
Hi,
I tried to discover novel junctions by tophat. One of the junctions it reported looked like this:
--In junctions.bed:
chr10 101101827 101102012 JUNC00036836 1 + 101101827 101102012 255,0,0 2 35,46 0,139
--On UCSC (see attached image):
My question is:
This appeared to be a false positive, though there were much lower coverage in the "intron" and the splice sites were canonical "GT-AG". There was only one alignment covering the junction (I checked the alignments by manually looking into this region of the SAM file).
Given the default -F/--min-isoform-fraction 0.15 and the high coverage in this region, should this junction not be considered? It seems like this is also an example of tophat looking into a single island for junctions as described in the tophat paper, otherwise the exon-exon junction would not be assembled in the first place.
Can someone explain a little bit how this --min-isoform-fraction work and why it is the case in my example?
By the way, is the 5th column in junctions.bed a coverage score?
Many thanks!
I tried to discover novel junctions by tophat. One of the junctions it reported looked like this:
--In junctions.bed:
chr10 101101827 101102012 JUNC00036836 1 + 101101827 101102012 255,0,0 2 35,46 0,139
--On UCSC (see attached image):
My question is:
This appeared to be a false positive, though there were much lower coverage in the "intron" and the splice sites were canonical "GT-AG". There was only one alignment covering the junction (I checked the alignments by manually looking into this region of the SAM file).
Given the default -F/--min-isoform-fraction 0.15 and the high coverage in this region, should this junction not be considered? It seems like this is also an example of tophat looking into a single island for junctions as described in the tophat paper, otherwise the exon-exon junction would not be assembled in the first place.
Can someone explain a little bit how this --min-isoform-fraction work and why it is the case in my example?
By the way, is the 5th column in junctions.bed a coverage score?
Many thanks!