Advice on RNA-Seq Depth

[KDS]

I am planning my first RNA-Seq experiment and wanted to make sure I was on the right track. I am going to be sequencing the transcriptome of 700 human lymphocyte samples on the Illumina HiSeq 2000 to look at expression differences, alternative splicing, and SNPs.

After looking through numerous papers, it doesn't seem like there is a clear consensus on the level of coverage needed. It seems like the majority of papers recommend a minimum of 30X coverage. Does anyone have any input on going lower or higher than 30X coverage?

Thanks for your help!

[tonybolger]

I am planning my first RNA-Seq experiment and wanted to make sure I was on the right track. I am going to be sequencing the transcriptome of 700 human lymphocyte samples on the Illumina HiSeq 2000 to look at expression differences, alternative splicing, and SNPs.

I'm not sure you really want to use the same library / sequencing strategy for both differential expression and alternative splicing/SNP detection. For DE, you want the most reads possible, and you get diminishing returns above 40 or 50 bases, while for alternative splicing / SNP detection you want long reads, especially if you're expecting novel transcripts, where normalized libraries, paired end and strand specific sequencing begin to matter as well.

After looking through numerous papers, it doesn't seem like there is a clear consensus on the level of coverage needed. It seems like the majority of papers recommend a minimum of 30X coverage. Does anyone have any input on going lower or higher than 30X coverage?

How do you even calculate coverage of RNA-Seq? You'd typically have some transcripts with >1000x, and a long tail with very low coverage, and probably 1000s you miss entirely.

[woodydon]

You should calculate the sequencing depth of a gene or an exon. It doesn't make sense to have a number representing the overall sequencing depth of RNA-Seq. DNA and RNA-Seq are different.

Besides, it is hard to believe that you will be sequencing 700 samples without knowing the differences of DNA-Seq and RNA-Seq.

Finally, your budget determines how you do sequencing. ~$100 per sample or ~$10 per sample will have entirely different designs. The more you spend, the more information you get. It is that simple. Some people would say 5 ~ 10 million reads is enough for RNA-Seq. That is totally bull**** and nonsense.

Don't bother the coverage illusion. Go find out how much money you are going to spend one a sample.

[puggie]

700 samples sounds like a lot, guess you'll be pooling many samples on the lanes unless you have good resources, one lane would give something around 400 million reads. Are you looking for anticipated highly differentially expressed genes from priori knowledge?