I'm setting up my first RNAseq experiment and I'm wondering what the opinion is of those with more experience. I have 50 samples, 25 from the control group and 25 from the treatment group. I'm really only interested in determining if a handful of immune genes are upregulated in my treatment group. Is 12 the max # of samples that can be multiplexed in a single lane? If so, how do I determine what my coverage would be for each sample? How do I determine what the coverage is for a different number of samples (X) on a lane? Right now, I'm thinking that I can put 6 samples on a lane, which would get me 16+ mil reads per sample, then just run 8 lanes. Is that correct? Is 16M reads sufficient for this type of question?
In the spirit of Devon, when discussing coverage and sequence depth, always specify species. I think about your problem somewhat differently: the maximum number of samples that can be multiplexed in a lane is determined by the number of uniquely barcoded libraries you can make. For instance, if you have access to 50 bar codes, you could make a pool of all your samples, and then run as many lanes as you need for a given amount of depth. Each lane would contain all samples. When analyzing the resulting data, you have to reassemble results, i.e. if all 50 samples were run on each lane, and you ran 8 lanes, you have to combine 8 results files for each sample. This is not an uncommon scenario. Advantages include thinking more freely about how many reads you need to answer a given question, less worry about separating experimental factors between lanes (i.e. all controls in one lane, all treatments in the other). Disadvantages include having to pool results at the end - but this is trivial with a script. If you need more depth, you can always run more lanes on your pools later. There are many ways to think about depth, multiplexing, and lanes.
Assuming that you'll be running things on a HiSeq (always specify the type of machine), then 12 samples on a lane will give you ~16 million reads/sample (you get ~200 million single-end reads per lane). That's probably a good amount of depth for the samples you have and then you could use ~4 lanes. Yes, you can multiplex more samples per-lane than that, but ~10 million reads is a common minimum recommendation for RNAseq.