Speaking of odd ideas that might have applications in drug discovery, there's an interesting one in the latest issue of Nature Methods (2, 31). A group at the Molecular Sciences Institute in Berkeley reports a new way to detect and quantify molecular binding targets. And if you think that this sounds like something we're interested in over here in the drug discovery business, you are correct-o-matic.

This idea piggybacks, as you might expect, on the mighty king of detection and quantification in molecular biology, PCR. The ability to hugely amplify small amounts of DNA is unique, the biochemical equivalent of a photomultiplier , and many people have taken advantage of it. In this case, they also make ingenious use of weird beasts called inteins, about which a great deal of background can be found here. Briefly, inteins are sort of DNA parasites. They insert into genes and are read off into an extraneous stretch of protein in the middle of the normal gene product. But then they quickly clip themselves out of the protein - they have their own built-in cut-and-splice mechanism - and leave the originally intended protein behind them, none the worse for wear.

The MSI group takes the molecule of interest - say, a protein ligand - and attaches an intein to it. They take advantage of its splicing mechanism to have the intein remove itself and attach a stretch of specially whipped-up DNA, which serves as a tag for the later PCR detection. They call this conjugate a "tadpole", for its shape in their schematics (the DNA tag is the tail, naturally.) Said tadpole goes off and does its thing in the assay system, binding to whatever target it's set up for, and you do a PCR readout.

The paper demonstrates this in several different systems, going all the way up to a real-world example with blood serum. What's impressive about the technique is that it seems to work as well as antibody methods like ELISA. Getting a good reliable antibody is no joke, but these folks can make smaller proteins with much worse intrinsic affinity perform just as well. And if you turn around and do the trick starting with an antibody, you can increase the sensitivity of the assay by orders of magnitude. And you get a real quantitative readout, with about +/- 10% accuracy. To give you the most startling example, the authors were able to detect as few as 150 single molecules of labeled bovine serum albumin in a test system.

The "News and Views" piece on all this in the same issue points out that the technique gets round some real problems with the existing methods. Labeling proteins with DNA or fluorescent tags is a messy and imprecise business, and it can be very hard to tell how many labels your target carries (or how many different species are really present in your new reagent.) The intein method is one-to-one label-to-protein, with no doubts and no arguing. Cell biologists are going to have to get used to knowing what they're looking at, but I think that they'll be able to adjust.

The news article calls the technique "ultrasensitive, amplified detection of anything," and that's pretty close. As the MSI authors point out, it removes the limitations of antibody technology: no longer can you detect only the things that an immune system has a reaction to. Screening of protein libraries could provide low- to medium-affinity partners (which is all you need) for all kinds of poorly-studied molecules.

I'd be interested in seeing if the system can be adapted for small (i.e., drug-sized) molecules conjugated to DNA. They wouldn't be tadpoles any more, though - more like eels - and might behave oddly compared to their native state. But even if you stick with the larger protein molecules, important biology may well be a lot easier to uncover. And we've got an endless appetite for that stuff. It's good news.