Tophat different outputs according to the different number of threads

[kumtl]

Hi! All,

I don't know anyone who had the same issue like I had, while using tophat software. I have a server which has 16 cores.

I tried to run tophat with different thread number for one of my sample. Unfortunately I had different outputs. (single-end RNA-seq data)

1. tophat -p 16 --no-coverage-search --solexa-quals --keep-tmp --b2-very-sensitive -G Gene_GTF --no-novel-juncs -o Tophat_out bowtie_index 1st_sample.fastq

2. tophat -p 4 --no-coverage-search --solexa-quals --keep-tmp --b2-very-sensitive -G Gene_GTF --no-novel-juncs -o Tophat_out bowtie_index 1st_sample.fastq

Any idea?
Thanks in advance...

[simonandrews]

When you say you got different outputs do you mean that you got the same hits, but in a different order (which wouldn't be surprising), or that you got different hits (which would)?

[kumtl]

simonandrews, thanks for your inquiry

Of course, I know about different order of outputs.
In my case, the result hits were different. Simply, the read numbers in accepted_hits.bam and unmapped.bam were slightly different. (-p 4 vs -p 16)

I don't know why this problem was happened. I tried the other number of threads (like, 1, 2, and 4) and their results were all same, except 16 (max num of available threads).

--Kumtl

[simonandrews]

If this is reproducible then I think it's one to send to the tophat authors.

[kumtl]

That's not bad idea. I'll let them know, if this problem happened again.
Thanks

[kumtl]

Simon, I have another question about tophat.

As you know from my post, I had different outputs because of different number of threads.
So, I tried to remap reads from unmapped.bam, and I got another hits from them. Of course, I applied exactly same option on tophat command line.

Do you have any idea?