Biological Replicate for RNASeq ?

[dhanapala]

• What will be the minimum number of biological replicates (plant samples) and how many biological replicates are needed for RNA-Seq using HIseq ?

[jwfoley]

Depends on your experimental design, but absolutely not less than two per experimental condition. It doesn't matter which platform you use.

[cascoamarillo]

One more question about the "replicate" definition; if you have a cDNA from a biological sample (for a RNA-seq experiment) and you use it into two PCR reactions with different barcodes (like the one you do before the Illumina run), the barcode1 and barcode2 samples from the same cDNA can be considered as "replicates".
Suppose you want statistical power in your data, and you want to test differential gene expression in programs like DESeq (where it wants replicates to check the so called "shot-noise"); in this case, can they be considered good "replicates"?

Thanks

[SamCurt]

One more question about the "replicate" definition; if you have a cDNA from a biological sample (for a RNA-seq experiment) and you use it into two PCR reactions with different barcodes (like the one you do before the Illumina run), the barcode1 and barcode2 samples from the same cDNA can be considered as "replicates".
Suppose you want statistical power in your data, and you want to test differential gene expression in programs like DESeq (where it wants replicates to check the so called "shot-noise"); in this case, can they be considered good "replicates"?

Thanks

These are more like technical replicates to me...

[jwfoley]

Yes, these replicates are only testing the reproducibility of the sequencing protocol, which is extremely high. Biological reproducibility is what we're interested in. I would just pool the data from technical replicates as a single biological replicate.

[chadn737]

I would do no less than three independent biological replicates (i.e. RNA isolated from separate biological samples) of each condition. You can get meaningful results from 2 biological reps, but you are really restricting yourself by doing this, particularly on a HiSeq where you can easily multiplex and still get a lot of reads per sample even on a single lane.