TopHat v1.1.4 - Empty *.bed files

[lianzinho]

Hello folks, I´m trying to run tophat on some 27bp single end illumina reads with a reference GTF for searching splice junctions.

My mapping reference is just the chr14 and so is the GTF that I retrieved from UCSC genome browser.

The problem is that in despite of not showing any errors in the log files, my *.bed output with the junctions is always empty!! So I´m in doubt if there is something wrong of if there really wasn´t any output what I find a little odd bacause of the long time it took to try to align the reads with the junctions. Take a look at the log file of the output:

[Fri Nov 26 08:19:34 2010] Beginning TopHat run (v1.1.4)
-----------------------------------------------
[Fri Nov 26 08:19:34 2010] Preparing output location /usr/local/tophat//
[Fri Nov 26 08:19:34 2010] Checking for Bowtie index files
[Fri Nov 26 08:19:34 2010] Checking for reference FASTA file
[Fri Nov 26 08:19:34 2010] Checking for Bowtie
       Bowtie version:                  0.12.7.0
[Fri Nov 26 08:19:34 2010] Checking for Samtools
       Samtools version:                0.1.11.0
[Fri Nov 26 08:19:34 2010] Checking reads
       min read length: 27bp, max read length: 27bp
       format:          fastq
       quality scale:   phred33 (default)
[Fri Nov 26 08:21:25 2010] Reading known junctions from GTF file
[Fri Nov 26 08:21:49 2010] Mapping reads against chr14.ucsc with Bowtie
[Fri Nov 26 08:26:23 2010] Joining segment hits
[Fri Nov 26 08:29:13 2010] Mapping reads against chr14.ucsc with Bowtie(1/3)
[Fri Nov 26 08:34:02 2010] Mapping reads against chr14.ucsc with Bowtie(2/3)
[Fri Nov 26 08:38:43 2010] Mapping reads against chr14.ucsc with Bowtie(3/3)
[Fri Nov 26 08:43:27 2010] Searching for junctions via segment mapping
[Fri Nov 26 08:43:31 2010] Retrieving sequences for splices
[Fri Nov 26 08:43:35 2010] Indexing splices
[Fri Nov 26 08:43:35 2010] Mapping reads against segment_juncs with Bowtie
[Fri Nov 26 09:59:20 2010] Mapping reads against segment_juncs with Bowtie
[Fri Nov 26 11:11:12 2010] Mapping reads against segment_juncs with Bowtie
[Fri Nov 26 12:25:22 2010] Joining segment hits
[Fri Nov 26 12:31:28 2010] Reporting output tracks
-----------------------------------------------
Run complete [04:12:35 elapsed]

Aparently it´s ok, but if I look to the junctions.bed it´s shows only the title:

track name=junctions description="TopHat junctions"

Please help, tophat parameters are default, i just changed the -p option to use all the cpus of my machine.

Thanks,
Halian

[wdt]

Hello there,

I am having a similar issue with empty junctions.bed file and was wondering
if you were able to resolve it successfully. Could you
share the solution?

Thanks,

[CompBio]

Were either of you able to solve your problems? I'm running into the same thing with TopHat 1.3.0 -- all the .bed files are empty, save for the file headers. Currently I'm running using default options with the following exceptions:

-a 10
-g 1
-F 0

In desperation I may drop the anchor size to something silly, like 4, or raise mismatches to 2 or 3. But I was able to align the same data with an old version of TopHat with these same parameters.

[jjpurwar]

Im having the same issue. Interestingly, when my input file has reads that are 50 bp, my *.bed files are NOT empty. However, when using a 36 bp input file, the *.bed files are empty.
Any clues on how to fix this?

[jjpurwar]

Figured a fix for this - the segment-length needs to be set to atleast half of the read length. else junctions wont be reported.

[thurisaz]

jjpurwar, do you mean to say that segment-length should be at most half the read length (or, alternately, that the read length should be at least 2 * segment-length)? That would seem to make more sense given the description for --segment-length in the manual:

"Each read is cut up into segments, each at least this long. These segments are mapped independently. The default is 25."

As I understand it, if segment-length > 0.5 * read-length, then only one segment will be made for each read (and the remainder discarded?), which is why it's having trouble finding junctions.

I got an empty junctions.bed (and a "junction database is empty!" message) when trying out TopHat with the default settings on some published RNASeq data with 35-bp SOLiD reads; I'm about to try a new run with segment-length set to 17 to see if that works.