Oh, a talk, erm...

So I’m writing this paper on a project that my boss imposed on the laboratory. As you might know, because I’m a complaining sort of blogger at times, and I’ve wailed about this thing before. The idea was to go fishing for mutations with our fancy new sequencer in a kind of tumor whose origins remain mysterious. We couldn’t cover the whole ocean of possibilities, so we stocked the pond with a reasonable list of candidates, and indeed hauled out a bunch of mutations. Six genes times five tumor characteristics isn’t that much analysis to do.

.
Over the summer, I sent around a draft of the paper, and it came back that it might be very useful to complement the mutation information with info on the expression of some pretty common markers that interact with the genes we found mutated. That would tell a more complete story, and the work wouldn’t take long. Just five protein markers.
Five isn’t too many.

But every time you add a marker, you need to compare it to the results of each one of the others. The amount of number crunching to do goes up exponentially, and the different possible combinations of results far exceeds the number of tumors we had to look at.

.
Fine. I did all that analyzing and hoped to find the new data neatly confirming what we were learning from the mutations and consolidating the tumors into a small number of groups with similar biology. Getting a handle on the biology might give us new therapeutic options.
Only, no, that didn’t happen. For every two the same, one was different, any way you looked at it.

.
It was like thinking that (kilt + bagpipe) might = Scots, and then assessing haggis just to be sure.
Well: kilt could be Scot or Halloween, bagpipe could be Scot or Breton, and haggis could be Scot or just weird. And none of the bona fide Scots are required to have kilt or bagpipe or haggis at all!
Things just fall into chaos when you don’t have enough samples for the p values to get really small.
**Sigh**

So the new paper has been written and sent around with a plea for advice (and not just my awful English, please!), and after a month only one of my faithful coauthors has responded. She, alas, seems to have dug up the old version to advise me on.

Meanwhile, the guy whose idea this all was in the first place remains silent. Oh, except to schedule me for a public talk on this very project a mere three weeks from now.

Thanks. I’m honored.