Linda threatens Amanda, 1007
Amanda is not impressed
Okay, I know that I implied (in my last post) that you were rid of me until the end of August. However, I am at the Hutch, waiting for a phone call, and I am just not up to attempting another research paper. So, I am going to write about some idle musings I had while riding the shuttle yesterday. They concern some differences between the kind of science I used to do and what is done – well, a good part of what is done – around here.
The first difference is the sheer size of data sets. The largest study I ever published described the paleomagnetism of the Sanpoil Volcanics, a ~ 50-60 MY old sequence of andesites in eastern Washington. Ken Fox and I were P.I.s, and we had two graduate students. We sampled 90 sites, each with 5 or so samples. Each sample was cut into two or three specimens – but usually only one specimen per sample was measured. Demagnetization procedures required, say, an average of 5 measurements per specimen. Thus, in all, about 90 X 5 X 5 = 2250 measurements were made, although only site-means (90) were reported. Thankfully, the graduate students did the measurements!
Contrast this with an experiment I just ran upon, described in Focus, apparently a newsletter for Harvard Medical. In this study, aimed at dealing somehow with the gene Myc, a well-know cancer troublemaker, nearly 75,000 things called shRNAs (short hairpin RNAs) were studied to find which were effective at “silencing” genes that Myc relies upon to do its dirty work. They found 403 potential candidates, and ultimately settled on one of them – coded for by a gene called SAE2 – for reasons that were not specified. Presumably work will proceed from there. One suspects that the 75,000 measurements were made by some sort of expensive “high through-put” machine and not graduate students, who are less reliable than machines and moreover have to be fed.
The second contrast between what I once did and what I am helping to do now concerns a fundamental difference in, for want of a better term, scientific philosophy. In our Sanpoil study we had deduced that dextral shear along the west coast of North America had affected continental crust fairly far inland. To test this hypothesis we sought to detect clockwise rotation of the Sanpoil block. The large data set merely allowed us to be extremely precise. By contrast – and given the usual caveats regarding my severely limited understanding – the Harvard shRNA Myc study seems largely inductive. They must have known that some types of shRNAs have the capacity to interfere with the oncongenic activities of mutated Myc, but they chose simply to test the entire kit and caboodle to find which ones. All 75,000 of them!
This is motivated by practicality; at least some aspects of cancer biochemistry seem to be so intricately involved that pure deduction just won’t work. In my geological example we already knew that relative motion between the North American and Pacific basin plates had generated shear in coastal North America; we simply wanted to see how far inland it extended. Piece of cake! But cancer research, it seems to me, is like one of those Russian Babushka dolls. Every time you succeed in opening one doll there is another inside to work on. Or two, or three, or – lots. Consequently, if we wait until we understand the biology well enough to deduce what might work, a lot of people will have died. So, God bless the through-put machines!
Keep on advocating.
No comments:
Post a Comment