biological replicate variation assessment

[neoanderson]

Hi to everyone,
I have rna seq data (bacterial) grown in rich and poor growth mediums. we have carried out 2 replicates of each of these conditions.

my question is, how do we computationally estimate how similar our biological replicate data is? what files will I use to determine the variability between two replicates? will I have to use the bam files (post alignment to a reference) or raw fastq files to ascertain this? is there a program that will take reads/alignments from the 2 replicates and tell me the percentage similarity between the two datasets?

I know this sounds simple, and I went through quite a few threads, but could not understand how to work this out. I am struggling.

would really appreciate any input.
Thank you,
N

[chadn737]

I assuming that in comparing similarity what you are really interested is comparing how similar gene expression is as opposed to something else like SNPs, etc.

For this you really need to extract read counts per gene and normalize the data. You can then use this data to cluster the samples or do a principle component analysis.

Take a look at the DESeq vignette or the EdgeR user guide which give nice descriptions of how to do both sample clustering and PCA for comparing samples:

http://www.bioconductor.org/packages/release/bioc/html/DESeq.html

http://www.bioconductor.org/packages/release/bioc/html/edgeR.html

[Lien]

Hi,

I would also like to know how similar my replicates are. I'm comparing RNA-expression in 2 different cell lines, with 3 replicates each.
I used the Tuxedo pipeline to get me a list of differentially expressed genes. However, I'm wondering if it is also possible to use this pipeline to estimate how similar my replicates are. When I look at my replicates in a heatmap, they look very similar.
I'm not very familiar with different pipelines, so therefore I'm wondering if it is possible with the Tuxedo-pipeline or CummeRbund?
Many thanks,
Lien