I have RNA-seq data under three different environmental experiments with the goal to detect differently splicing transcripts. When dealing with my three conditions, let's cond_x, cond_y, cond_z, I could:

• Strategy 1: run cuffdiff with all of their alignments as inputs
• Strategy 2: pair-wise compare.

I have tried both the strategies, but it turns out that only Strategy 1 would yield differently spliced events in isoform_exp.diff output file while strategy 2 yield nothing different.

Could anyone give me some information why this happen?

My command line is :

cuffdiff -o out_dir -L Cx,Cy,Cz --library-type fr-firststrand --num-threads 3 \
  --frag-bias-correct genome.fa ensembl.annotation.gtf \
  Cx_1.bam,Cx_2.bam,Cx_3.bam Cy_1.bam,Cy_2.bam,Cy_3.bam Cz_1.bam Cz_2.bam Cz_3.bam