Tophat junctions.bed

[RockChalkJayhawk]

Has anyone seen a wierd junctions file as an output. Although most of my paired-end reads aligned to the genome (both pairs), my junctions.bed file doesn't link together many of my exons. In fact, most of the links that were made are not even between exons, rather they were intergeneic and scattered.

This is not due to loading it onto the wrong assembly. Has anyone else had this problem with paired end reads?

[RockChalkJayhawk]

Also, the coverage.wig looks good.

[RockChalkJayhawk]

Here is a good example of what I'm talking about. For this gene I have a lot of coverage, but only 3 splice junctions are detected. Any ideas?

[lifeng.tian]

Originally Posted by RockChalkJayhawk

Here is a good example of what I'm talking about. For this gene I have a lot of coverage, but only 3 splice junctions are detected. Any ideas?

I looked at my TopHat results for a 75nt run (18M paired-end reads, illumina GA IIx), there are 12 junctions detected in chr1:879,584-894,679 region, however, only 4 junctions have more than 5x coverage. If you have shorter reads and lower coverage than
my data, you may see less number of junctions here.

FYI, I detected 185,238 total junctions.

Here is the 4 junctions from my 75nt run:
chr1 887451 887857 JUNC00121200 40 - 887451 887857 255,0,0 2 68,66 0,340
chr1 889202 889440 JUNC00121202 37 - 889202 889440 255,0,0 2 70,57 0,181
chr1 891331 891529 JUNC00121205 17 - 891331 891529 255,0,0 2 62,55 0,143
chr1 894390 894642 JUNC00121207 7 - 894390 894642 255,0,0 2 71,48 0,204

[john_mu]

Originally Posted by RockChalkJayhawk

Has anyone seen a wierd junctions file as an output. Although most of my paired-end reads aligned to the genome (both pairs), my junctions.bed file doesn't link together many of my exons. In fact, most of the links that were made are not even between exons, rather they were intergeneic and scattered.

This is not due to loading it onto the wrong assembly. Has anyone else had this problem with paired end reads?

This would depend on the settings you use. SNPs in the right location could cause the junction search to fail.

What do you mean by most? if you mean over 50% of your gene exhibit that kind of behavior despite good coverage that is a bit strange. Maybe you should try less strict settings.

or you could give SpliceMap a try

[RockChalkJayhawk]

Originally Posted by john_mu

This would depend on the settings you use. SNPs in the right location could cause the junction search to fail.

What do you mean by most? if you mean over 50% of your gene exhibit that kind of behavior despite good coverage that is a bit strange. Maybe you should try less strict settings.

or you could give SpliceMap a try

I thought about it, but my reads are 2x36 -- too short for SpliceMap. I'm still playing around with some of the parameters...

[edge]

Hi lifeng.tian,

I'm running tophat with the following command:

tophat -r 110 -p 4 --solexa1.3-quals --library-type fr-unstranded -G homo_sapiens.gtf human_reference_genome s1_1.fq s1_2.fq

OUtput file generated from tophat:

junctions.bed
insertions.bed
deletions.bed
logs/
accepted_hits.bam

Based on the junctions.bed, can I know that how you group them based on the region in each chromosome?
This is the total number of junction identify by tophat:

chr1    13677
chr10   5130
chr11   7780
chr12   7558
chr13   2293
chr14   4818
chr15   4368
chr16   6604
chr17   8343
chr18   1778
chr19   8369
chr2    9800
chr20   3278
chr21   1460
chr22   3419
chr3    7619
chr4    4735
chr5    5970
chr6    6789
chr7    6513
chr8    4306
chr9    5461
chrX    4042
chrY    312

What is the important to identify its coverage at this stage?
Is it the figure in column 4 represent the coverage of each junction?
Really thanks for any of your advice to interpret the junctions.bed shown in tophat_out.
Many thanks in advance.

[arcolombo698]

Originally Posted by lifeng.tian

I looked at my TopHat results for a 75nt run (18M paired-end reads, illumina GA IIx), there are 12 junctions detected in chr1:879,584-894,679 region, however, only 4 junctions have more than 5x coverage. If you have shorter reads and lower coverage than
my data, you may see less number of junctions here.

FYI, I detected 185,238 total junctions.

Here is the 4 junctions from my 75nt run:
chr1 887451 887857 JUNC00121200 40 - 887451 887857 255,0,0 2 68,66 0,340
chr1 889202 889440 JUNC00121202 37 - 889202 889440 255,0,0 2 70,57 0,181
chr1 891331 891529 JUNC00121205 17 - 891331 891529 255,0,0 2 62,55 0,143
chr1 894390 894642 JUNC00121207 7 - 894390 894642 255,0,0 2 71,48 0,204

for this file (in response to #4) how did you calculate the total amount of junctions? is it the total amount of entries in your junctions.bed file? ie. the amount of columns in your file?