[HESmith]

If you already have the sequencing done you obviously can't do this, but adding spike-in controls would go a long way to letting you compare expression levels across datasets. Also, biological reps.

Actually, no. Spike-ins are only useful for validating the reproducibility of library construction (i.e., technical replicates). They tell you nothing about how much transcript levels vary b/t samples (biological replicates).

[danwiththeplan]

Actually, no. Spike-ins are only useful for validating the reproducibility of library construction (i.e., technical replicates). They tell you nothing about how much transcript levels vary b/t samples (biological replicates).

Well, my understanding was that spike-ins (ERCC spike ins like this: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3166838/) allow you to control for effects that would otherwise go undetected between runs (gc content, length effects, concentration effects), whether those runs are biological or technical reps, and that spike-ins would allow you to detect these effects from the point that they are spiked in (which should be as soon as possible), so, they are useful for validating all of the steps from that point onwards (i.e. polyA purification, ribodepletion, library construction). But maybe I misunderstood and I'm open to being corrected.

Plus, I'm not sure why anyone would do technical reps anyway when there's so much more value to be gained from another biological rep.

[HESmith]

Okay, my answer was clearly too terse. When RNA-Seq data indicate a difference in gene expression, there are two potential causes: biological (i.e., transcript levels go up or down in the cell/tissue/organism), and technical (changes introduced on the path from nucleic acid isolation to sequence generation). Spike-in controls are useful for detecting/validating the points you mention, but those are all technical aspects of the experiment. Spike-ins do contribute to the biological basis for the observed difference, but only in the sense of eliminating technical bias as the culprit.

[danwiththeplan]

Okay, my answer was clearly too terse. When RNA-Seq data indicate a difference in gene expression, there are two potential causes: biological (i.e., transcript levels go up or down in the cell/tissue/organism), and technical (changes introduced on the path from nucleic acid isolation to sequence generation). Spike-in controls are useful for detecting/validating the points you mention, but those are all technical aspects of the experiment. Spike-ins do contribute to the biological basis for the observed difference, but only in the sense of eliminating technical bias as the culprit.

Yep. Well put.