Microfluidics: Miniature Chemistry Labs

enkidu writes: "The NYTimes has a story (free reg, yaba yaba) about the rapidly emerging field of microfluidics and describes some of the methods used in making micro-valves, pumps and other components. In the future, you won't need to send your blood/urine sample to a lab, your doctor will put in his "lab-in-a-box" and hand you a printout before your leave."

Comments

Obligatory no-reg link

[no parity]

http://college.nytimes.com/2002/01/01/science/phys ical/01MICR.html

Incredible Shrinking Labs

[Anonymous Coward]

http://www.ornl.gov/ORNLReview/meas_tech/shrink.ht m
A small blurb on Dr ramseys work.

http://www.chipcenter.com/columns/bmcginty/col01 2. html
Another article on fludic chips.[check the links at the bottom]

Re:Won't this hurt accuracy?

[baz00f]

Speaking as a pedigreed biochemist, you are correct in the extreme of vanishingly small samples. But these devices are still working with enough volume such that an analyte of interest at a substantial concentration (glucose, cholesterol, etc.) is effectively present at the same concentration at nearly all sample volumes.

Things DO fall apart (as you intuit) when the concentration of the analyte gets vanishingly small. We see this routinely when we try to quanitate DNA using PCR (Polymerase Chain Reaction) methods. PCR is sensitive enough that we can detect ONE copy of a DNA molecule in a volume of sample. So if you have say, one copy in 1ml of volume, and you sample .1 ml and do your PCR, your test would come up negative upon repeats (on average) 9 out of 10 times. With small numbers of copies you can use Poisson statistics to calculate your hit rate. With higher concentrations your Poisson distribution collapses to a gaussian that gets narrower and narrower, which is the regime that most normal wet analytical techniques work. For example, fasting blood glucose is about 100 mg/dl, which is about 5 mM. Assuming your device can work with 1 nanoliter sample size (this is about 100x smaller than a volume about the size of the proverbial period at the end of a sentence) you would have 3x10^12 molecules of glucose in it. Assuming your technique is sensitive enough to register the presence of this "small" number of molecules, you are still far away from seeing sampling errors on repeats of the same sample due to random fluctuations of the number of molecules (the "concentration") in any given sample.

Paul Yager at U. Washington (Seattle) has a good introduction to microfluidics:

Microfluidics Tutorial and Prognostication

Re:Won't this hurt accuracy?

[Vinson Massif]

Analytical volumes are in the mL range to ease handling and to minimize cumulative error. In most modern techniques, microlitres are taken up by the test instruments; the actual amount of 'stuff' that is actually analysed is minute. Good test design and bench technique give good results.

It's not the sample volume I'd be questioning, but calibration, standarization, and QA procedures. These are the root of inaccurate (and imprecise) test results.

--