How many tags to perform a RNAseq experiment

[kant1_IC]

Dear All,

We planed to performed the same whole transcriptome analysis on SOLiD 5500xl and HiSeq 1000 using Human samples. We would like to know how many tags should we sequence to be able to determine fusion transcript and also to perform gene expression analysis.

Thanks in advance.
Best regards.

Kant1

[kopi-o]

Unfortunately, the answer is "as many as you possibly can" :-)

[kant1_IC]

that would not really help me because i already think the same !!!!

thanks for your help kopi-o

[kopi-o]

I know, but it's really not possible to answer your question with a specific number. See this blog post:

http://blog.fejes.ca/?p=607

[TonyBrooks]

There is no consensus here. It depends on what your overall goals are (finding novel transcripts, splice variants or DGE) and also on how complex your genome is.
Speaking from an Illumina perspective (I assume the same would hold true for SOLiD, they published a white paper saying that around 2m mapped reads shows similar sensitivity to a microarray (from "Vendor A") in terms of identifying the same % genes showing differential expression between MAQC brain and Universal Ref RNA.

http://www.illumina.com/documents/products/whitepapers/whitepaper_RNASeq_to_arrays_comparison.pdf

I've seen other people quote 10 m being equivalent to a microarray (probably also depends on the microarray).

Either way, we recommend >30 million reads per sample with at least 3 biological replicates. Ideally we would like more reads, but it's just that 30 million sits nicely in the niche of our what our sequencers can output (6 samples in one HiSeq lane, or 1 sample per GA lane).
I would also argue that biological replicates are more important than number of reads (i.e. 3 x 20m reads is better than 1 x 60 million).

[Seqasaurus]

35 million is my magic number

[kant1_IC]

I think that I got the answer that I was looking for.
I resume, I need at least 100 M of tags per samples (including 3 biological replicates).
Thank you guys

kant1