Cufflinks, two replicates have totally different expressions from cufflinks

[cloris816]

Hi everyone,
I'm doing mouse transcriptome analysis using Tophat-Cufflinks work flow, and I met a confused problem.
I have 8 samples, each of them has a replicate, so I have 16 samples of RNA-seq data. Then I remove adaptors using cutadapt. Tophat was performed with fastq data without adaptors, so I got a mapping rate around 80%. Tophat runs with default parameters. Then I use cufflinks on accepted_hits.bam files from Tophat (default parameters).
14 samples went well on this but 2. Let's name them A and A-rep.
21570 genes expressed in sample A, while only 364 genes expressed in A-rep. (FPKM value > 0 means gene expressed)

So I searched for the error file, and it gives me a warning
Warning: Using default Gaussian distribution due to insufficient paired-end reads in open ranges. It is recommended that correct parameters (--frag-len-mean and --frag-len-std-dev) be provided.

What's the meaning of this message? Because it went well on my other samples...

Many thanks!

[gringer]

You should be providing Tophat with your estimates of fragment length (--mate-inner-dist and --mate-std-dev). Even if it works without them, it's a good idea to include them because it can alter the speed and quality of results.

Everyone seems to have their own idea of the best statistic for describing the distance between reads. Here's Tophat's version:

This is the expected (mean) inner distance between mate pairs. For, example, for paired end runs with fragments selected at 300bp, where each end is 50bp, you should set -r to be 200. The default is 50bp.

[cloris816]

Thank you gringer! I finally got things done by adjusting parameters in cutadaptor not Tophat. The newer version of cutadaptor help removing adaptors for pair-end sequencing reads so that my reads are concodent to each other. thus Tophat identifies the reads successfully!