How to combine junctions.bed files produced by TopHat

[pengchy]

https://github.com/roryk/seqscripts has a bunch of little gluey type scripts to work with splice junctions, one of them does what you want. Download it and do this from the command line:

cat junctions_file1.bed junction_file_2.bed junction_file_3.bed| tbed2juncs | combineJuncs > combined.juncs

combine.juncs will be a BED file and contains all of the junctions from the junction files with the score the sum of the junction scores. The names of the junctions are the locations of the two nucleotides at the edges of the exons that are joined together.

great work! Thank you roryk.

[roryk]

Did you end up writing your own script that merges the junction.bed files from different sample while preserving the score? If yes, would you be willing to share that script since this is exactly what I need to do.

https://github.com/roryk/seqscripts has a bunch of little gluey type scripts to work with splice junctions, one of them does what you want. Download it and do this from the command line:

cat junctions_file1.bed junction_file_2.bed junction_file_3.bed| tbed2juncs | combineJuncs > combined.juncs

combine.juncs will be a BED file and contains all of the junctions from the junction files with the score the sum of the junction scores. The names of the junctions are the locations of the two nucleotides at the edges of the exons that are joined together.

[fongchun]

Did you end up writing your own script that merges the junction.bed files from different sample while preserving the score? If yes, would you be willing to share that script since this is exactly what I need to do.

[DrD2009]

I'm new just thought I'd try helping. Apparently it isn't the bed file format I am familiar with.

[HTS]

You could use Galaxy (http://main.g2.bx.psu.edu/). Upload the files (probably as file format tabular) you want to combine and then use the 'Concatenate queries' tool found under Text Manipulation. There's probably a short script out there to combine them, but this should work for what you want to do too.

-Brandon

Thanks for your reply, Brandon! I guess I didn't make myself clear but I need something considerably more sophisticated than concatenating multiple text files, especially considering what each field of the junctions.bed file means.

[HTS]

TopHat comes with an (undocumented, I realize) script called bed_to_juncs. Running it on a TopHat BED file will produce a .juncs-format file. To merge these, you can simply cat them together, sort, and pipe to uniq.

Hi Cole,

I am aware of the bed_to_juncs script but when converting BED files to .junc files, it doesn't preserve the score and overhang information that I would like to combine. BTW, could you confirm if my understanding of the score field in BED files is correct? Thanks a lot!

-- Leo

[Cole Trapnell]

TopHat comes with an (undocumented, I realize) script called bed_to_juncs. Running it on a TopHat BED file will produce a .juncs-format file. To merge these, you can simply cat them together, sort, and pipe to uniq.

[DrD2009]

You could use Galaxy (http://main.g2.bx.psu.edu/). Upload the files (probably as file format tabular) you want to combine and then use the 'Concatenate queries' tool found under Text Manipulation. There's probably a short script out there to combine them, but this should work for what you want to do too.

-Brandon

[HTS]

Hi,

Basically I would like to double check before writing my own script to do so. Since I need to pool together samples with different read lengths, I have to run TopHat separately for them (using the -j option wherever appropriate). I already know how to merge the resulting .sam files with samtools and generate a combined coverage.wig file. Before putting effort to combine the junctions.bed files, I would like to know:

1. If there are tools/scripts to do this already, either within TopHat or outside.

2. If none, I would like to confirm if the score field of the junctions.bed file is simply the number of uniquely mapping reads that are aligned to the junction, or if muitlreads are also counted/weighted.

Please feel free to share your knowledge/experience/comments. Thanks a lot!

-- Leo