tophat/Cufflinks workflow question

[hmortens]

I am still trying to make sure that I am running tophat/cufflinks in the most effective way, and was hoping folks might pass on how they set up their workflows.
I am testing four fastq files currently. I have run each fiule through tophat seperately, and only combine the files at the cuffdiff stage. It still seems like there should be a more effective way (aside from batching each sample) to analyze an entire flow cell for example.
I cannot get Tophat to align all four samples, trying mutiple variations to reads1_1.
So, we can go with seprate bam files to cuffdiff, and do seem to remember something about running cuffdiff on each sample seperately.
But, at what point do you then bring more than two samples together?
Any opinions, samples workflows that you use, or hints would be greatly appreciated!
thanks,
holly

[polyatail]

Are these four samples technical replicates (i.e. just different lanes on a flow cell), biological replicates, or different conditions? If they're technical replicates, it is reasonable to combine the FASTQ files prior to running TopHat. Biological replicates and different conditions you'd want to run separately and specify together at the Cuffdiff stage of the workflow.

Could you provide some more information about the reads? Are they paired-end? How are you preparing them for TopHat? What TopHat parameters are you using?

[hmortens]

I am using Joe Pickrell's data (http://eqtl.uchicago.edu/RNA_Seq_data/), which I believe are single end reads, and have been running them as such using the following settings.

> time tophat -r 200 -p 14 -o tophat_out_s_1 /illumina/runs/data/bowtie_indexes/hg19.ebwt.zip_FILES/hg19 s_1_sequence.fastq

so, I have been working with s_1, s_3, s_5, and s_7. each from different samples/individual.
From your message, I would continue to run each seperately through tophat.
At the Cufflinks stage, do you prefer to cuffmerge, and then run cuffdiff, or some other set of steps?
many thanks!