Friday, April 25, 2014

“Why don’t you just use [some old boring tech] instead?”

One common criticism/comment about measurement techniques that I hear scientists hurl at each other is “Why don’t you just use [some old boring tech] to measure that instead?” Like: “Couldn’t you just use RT-qPCR instead of RNA FISH?” Or “Couldn’t you get the same results with a western blot?” Or “Do you really need single cell analysis to get that result?” These points are fair ones, and often it is completely true you could use something less precise, cheaper, more standard, etc. to measure something. I sometimes say (or at least think) these sort of things myself.

But I think this ignores the fact that once a new, high resolution and more precise technique becomes standard, it can free your mind to think in new ways. Take an example from numerical methods (I forget who told me this example–I want to say it was my office mate at Courant, the absolutely brilliant Yoichiro Mori). When I type in sin(1.4787) into MATLAB, it just gives me the answer. It gives me the answer out to far more decimal places than probably need, but whatever, it just works. The most important part is that *if* what I’m doing requires more decimal places, I don’t need to think “oh, I should probably use this other algorithm for that, maybe that’s why this thing isn’t working”–no, it just works all the way, every time, whether I need it or not. Not having to worry about it frees your mind from the details and let’s you ask things you couldn’t or wouldn’t ask otherwise.

For a biological example, let’s say I need to measure a fold change in transcript abundance. In our lab, we’d probably do RNA FISH, which gives us absolute numbers and single cell counts, etc. Now, someone could say, well, do you need counts in single cells? Do you need absolute numbers? Why not just do RT-qPCR? Sure, fair enough. But to the extent that we believe RNA FISH is more accurate, then why not? Then I just don’t have to worry about all the controls, etc. And if I want to ask a question of my data that does require more accurate numbers, well, then I have it. The point is that I have the answer and I can move on to the next thing.

(Cost is of course an important consideration as well, and one that I'm not considering here. Costs change, though, and I also think that scientists often don't factor in the time required. RT-PCR is relatively cheap, but many fail to take into consideration the cost of the time used to validate primers, build standard curves, etc.)

I think this “arbitrary precision” principle is important for building assays that we can build upon. I think that it is generally true that it is hard to build new assays out of parts that are finicky and require a lot of careful consideration. Take Sanger sequencing. When we sequence, we don’t usually have to worry too much about how the loaded the sample, etc. It just works and we take it as such. Same with Illumina sequencing, for the most part. Or buying an off-the-shelf microscope. Yes, these methods and tools have important tricky points that we have to watch out for, but the point is that you don’t have to think about them most of the time. It makes it just that much easier to innovate further upon them. It's much harder to build something on top of a coin flip.

Anyway, just something to think about when someone introduces a fancy new method to measure something. Going to try and stay more open minded myself.

No comments:

Post a Comment