How to set Tophat insert size

[gringer]

Those are experimental values. You should have a good idea of your paired-end distance before doing any mapping. If you can get a good enough idea of that based on genomic (or transcriptomic) mapping distance by running bowtie, maybe you don't need to be running tophat at all....

[miaom]

I think the parameters (--mate-inner-dist, --mate-std-dev) are for Tophat only. Bowtie does not need to specify them.
Running bowtie can generate SAM files, which contain information about fragment size. Without pre-running bowtie (or maybe tophat), how can I know the values for --mate-inner-dist and --mate-std-dev to feed in Tophat?

[gringer]

Those are precisely the options that you apply to the tophat command line to change the expected insert size distribution for Tophat. Tophat takes as arguments the bowtie index (and optionally transcriptome GTF file) together with raw reads in FASTQ format, calling bowtie/bowtie2 internally. You shouldn't need to directly run the other programs in order to use Tophat:

Usage: tophat [options]* <genome_index_base> \
  <reads1_1[,...,readsN_1]> [reads1_2,...readsN_2]

Perhaps you could rephrase your question. What have you tried (command line is useful), and why do you think it doesn't give you the correct results?

[miaom]

correctly setting 2 parameters in Tophat:
--mate-inner-dist
--mate-std-dev
should be very important, but i have no idea how to set them without running bowtie (or bowtie2) -> samtools -> picard tools first, which seems a little bit redundant for Tophat.

Does any one have suggestions?
Thanks in advance!!