Need Help Designing Experiment.

[master_shake]

We have a human cell line that expresses a leukemia-associated fusion protein. I have created two cell lines that express one of two shRNA’s targeting this fusion protein, and one cell line that expresses shRNA to luciferase. The simplest experiment is to sequence the three cell lines (shRNA1, shRNA2, shRNA Luciferase), with the two fusion protein shRNAs serving as pseudo-biological replicates. We’d like to get the most information for the most reasonable cost, and with data that we can be reasonably comfortable with. We’d like to be able to compare this data to other data sets, usually microarray, which have been collected for this leukemia, as well as compare it to microarray data from mouse models. For this reason, we feel RNAseq may provide an advantage because it will cast a wide net over most genes, whereas doing this as a microarray experiment may miss some genes that we may later discover we are interested in. If the goal was to look solely at differential expression of coding transcripts, what would be a reasonable read depth to use (e.g., 2x75 Paired end at 30 million reads)? If the goal is to get more information about noncoding transcripts as well and maybe alternate splicing (i.e., more fully taking advantage of RNAseq), how ought we to change the parameters of the sequencing experiment? I realize increasing the number of biological replicates is always ideal, but before undertaking a massively expensive experiment that may yield no usable data, I need help in being able to think of how to best devise the initial sequencing strategy. It may be that I do a pilot RNA seq experiment with the three samples that gives us information about many transcripts, and later do follow ups with a cheaper technology, such as microarray.