Gene expression analysis-comparsion

[bioman1]

I have normal tissue (control A, control B) and abnormal tissue (treatment A, treatment B). I need to do differential gene expression analysis. How can I get up regulated and down regulated genes occur in normal tissue and abnormal tissue?. What are the steps to followed. The below are my steps, I think my procedure bit confusing, please better way to compare?

1. I have transcript of normal tissue (control A ,control B) and replicate tissue (treatment A, and treatment B).
2. I have denovo assembled separately (control A, control B, treatment A and treatment B transcriptome) and read mapped each reads to denovo assembled to each one
3.Got FPKM value for each control A, control B, treament A and treatment B.
4.How to get up regulated gene and down regulated gene for comparing normal and abnormal tissue?

[Brian Bushnell]

You can't get valid differential FPKMs by mapping reads to their own denovo assembly; they all need to be mapped to the same reference.

[bioman1]

Do I need do denovo assemble all samples (control A, control B and treatment A and treatment B) all in one reference?
Then how can I compare to control A, treatment A and control B, treatment B and get up-regulated and down regulated genes?

[mikep]

You don't denovo assemble anything.

You map the 4 samples to the same reference genome using a mapper like BWA, bowtie, or if you are sequencing something that alternate splices a mapper like tophat or rna-star.

You then use some kind of gene counter like htseq-count to determine reads per gene, and finally you use something that determines differential gene expression like DESeq.

Having said that, you only have duplicates, so any statistics generated will be dodgy.

[bioman1]

I don't have reference genome and also I dont know close reference genome.
Can I do denovo assembly of transcripts reads (control A, control B, treatment A and treatment B)?
Map each reads (controlA.fastq, controlB.fastq, treatmentA.fastq and treatmentB.fastq) to denovo assembled transcriptome?.
Getting mapped reads and do gene expression analysis?. Then how can i compare (control A, treatment A to control B and treatment B)?. Get up-regulated and down regulated genes?

[mikep]

Ummm.... you are in trouble.

I'd dump all the reads from all experiments into 1 big file then do a de novo assembly.

Then turn that into your reference genome, and map each sample back against it.

If you are lucky you'll end up with a bunch of differentially expressed contigs, which you can then blast to determine what genes they are likely to represent.

Alot depends on the complexity of the organism.

[Brian Bushnell]

I'd dump all the reads from all experiments into 1 big file then do a de novo assembly.

Then turn that into your reference transcriptome, and map each sample back against it.

If you are lucky you'll end up with a bunch of differentially expressed contigs, which you can then blast to determine what genes they are likely to represent.

Agreed; that's what I'd do.