RNA-seq triple replicates

Question

[biolab]

Dear all,

I have triple replicates of RNA-seqs for both control and mutant samples. All of these six seqs are single-end un-stranded RNA-seqs. I ran tophat2 for all six RNA-seqs, but then do not know how to run cufflinks, cuffmerge and cuffdiff commands. Can I merge three control and three mutant seqs? Thank you very much for any of your suggestions.

Answer 1

[Devon Ryan]

You run cufflinks on each of the 6 samples separately, run cuffmerge once with all 6 GTFs, and then run cuffdiff once with the results.

[biolab]

Hi, Devon Ryan, your answer is really helpful. I just have one more question: previously I ran cuffdiff using the following command, however this time I have six bam files rather than two bam files, could you please provide me a cuffdiff command? Thank you very much!

cuffdiff -o wt_mut_cuffdif -b ~/at/TAIR10.fas -p 4 -L wt,mut -u merged_asm/merged.gtf wt_tophat/accepted_hits.bam mut_tophat/accepted_hits.bam

[Devon Ryan]

cuffdiff -o something -b foo.fa -p 4 -L WT,MUT -u some.gtf WT1/accepted_hits.bam,WT2/accepted_hits.bam,WT3/accepted_hits.bam MUT1/accepted_hits.bam,MUT2/accepted_hits.bam,MUT3/accepted_hits.bam

Note that there are no spaces between the sample names and commas within a group, but the groups themselves are separated by a space. Adding aberrant spaces is a very common mistake to make at this point.

[biolab]

Hi, Devon,

This command and your description are really really helpful! Thanks a lot.

Answer 2

[Evgeniia Golovina]

Hi,

1. yes, you can merge three control and three mutant samples - in order to get 2 bam files. But it's optional step:

   samtools merge <out.bam> <in1.bam> <in2.bam> [...]
   

2. to calculate isoform abundance, you need to quantify FPKM (Fragments Per Kilobase of exon per Million fragments mapped) values:

   cuffquant -u <annotation.(gtf/gff)> <aligned_reads.(sam/bam)>
   

3. And you can perform analysis between two groups of samples. Actually, you can specify all your 6 samples in cuffdiff command. Look more about cuffdiff command here:

   cuffdiff <other options> --labels conditionA,conditionB <annotations file> sample1.bam,sample2.bam sample3.bam,sample4.bam
   

[Devon Ryan]

It should be noted that (2) should not be done on the output of (1), but rather on the unmerged files.

[Evgeniia Golovina]

Yes. You're right. Thank you!