Okay. (Ben slowly gathers thoughts...)...
Now I'm going to tell you about a paper that is so cool that even though I have no idea how they did it, I still think its worth sharing. I'm hoping I'll figure it out as I write this.
First of all, its Open Access (yay!) and available here! Second of all, its cool enough that 2 people sent it to me since it came out and this morning I thought I'd get it on the second read through.
What I do get: The Cancer Genome Atlas is not a leather bound book that sits in a room that smells of rich mahogany....
...instead, it is a huge cohort of clinical cancer samples that have have been or are in the process of being studied with a ton of different genomics techniques. The homepage of the project is here.
Browsing through the papers that have been done on this Atlas (to construct this Atlas? that makes more sense...) shows that there is a lot of bioinformatics firepower at work here.
So...in this study this group took these samples and did an interesting protein array analysis of them. This is where I get foggy. The array they used is called an RPPA. This is a Reversed Phase Protein Lysate Microarray (wikipedia link) (and if are a Jove user, or care enough to register for a free trial, here is a video that shows how an RPPA works.)
Okay. So they are using fancy antibody arrays to show the presence/absence/abundance of proteins. Got it. What do the arrays detect? Well, they went for a whopping 181 antibody probes! Wait? What? Just 181 targets? And the targets were selected based on what we know of current cancer pathways and stuff. My assumption is that the arrays are very fast and/or very cheap...or we would have done this with a mass spec and looked at hundreds of targets with PRM (people are routinely doing 700+ per assay these days on Q Exactives) or more with SRM, right?
But this is where it gets impressive -- monitoring all 181 targets on these arrays they looked at over 3,000 different samples...which is a lot... And these samples have been previously clustered by neat things like disease type and primary driving mutation. So, you can see how different genes interact with hundreds of samples of the same disease that follow the same -- or different cancer driving pathways.
Take home point for me is: For you guys out there generating insane amounts of clinical data, we need to steal more genomics tools! Cause these guys seem (at least...to an outsider...) to be able to do stuff with the data!