Hi all
I have been asked to identify genes that are specific to the suprachiasmatic nucleus (SCN) in mice. The RNA-seq datasets I have access to are from many different tissues. One of the data sets is from SCN tissue itself. Another of the data sets is from the brain. Which part of the brain you may ask? Recent papers on identifying tissue specific genes identify brain as a single tissue (here, here, here, and here). It is important to include brain as it has been identified as a tissue that has a large proportion of tissue specific genes.
The ideal experiment would be to compare genes found in SCN tissue to genes found in many other tissues including brain tissue that has had the SCN excised. But in the time frame I have it isnt possible to get data from the brains of mice that have had the SCN removed.
Because the SCN is part of the brain..... any genes found in SCN will be found in brain tissue. So how do I identify SCN specific genes?
The only idea Ive had to do this properly is to compare SCN tissue read counts with brain tissue read counts. If gene 1 produces X amount of reads in SCN tissue, and the same amount in brain tissue, then the SCN tissue accounts for all of the reads in brain tissue. Therefore it is SCN specific. The big catch though is that tissue specific genes tend produce a low number of reads. So I will be comparing very small numbers...that may not be statistically reliable (Im not a stat head but Im aware statistics tend to be important!)
Thanks in advance for the help.
Kenneth.
Knowing more than a bit about the brain in general and the SCN in particular, I'd be a bit surprised to find any particular gene truly specific to the SCN. The knee-jerk reaction would be to check core-clock genes, but I bet you'll find them everywhere. You might instead need to find a combination of things that together act as a marker.
Anyway, on the technical side, find microarray datasets. Your assumption regarding the signals summing is true there (but not for RNAseq). This is probably what I'd try too, though I'd also try to go through the Allen Brain Atlas, since they have nice histograms quantifying their in situ and other experiments by region and gene.
Thank you for the advice. A few points:
1) Find a combination of things that together act as a marker - I'll definately be discussing how this might be achieved with my supervisor.
2) Regarding microarray datasets - I am trying to avoid using microarray data because according to this review paper: "Our observations imply that past results which relied on microarray data for the evolutionary interpretation of tissue specificity, should be treated with great caution".
3) Im confused by "Your assumption regarding the signals summing is true there (but not for RNAseq)." - considering I'd prefer to use only RNA-seq data, is my assumption true or not?
4) Allen Brain Atlas - thank you for that! I had found a few other data sources (e.g. expression atlas) but not this one.
Thanks again.
I'll have to read the paper you linked to. Regarding point 3, signals from different tissues don't sum together in RNAseq, but they do in microarrays (to a point). That's pretty much the only benefit to microarrays these days.