Sequence-Detecting Nanoscale Sensor

Makarand writes "A nanoscale sensor made of a single molecule - just 20 nanometers long - capable of detecting a specific short sequence in a mix of DNA or RNA molecules has been created by physicists at UCLA. This nanoscale sensor could be used to detect the early stages of cancers for which genetic markers are well known or extremely minute traces of biological weapons. When a target molecule binds to the probe molecule in the sensor, the probe molecule changes shape and pulls on the sensor. The motion of the sensor is detected by an optical technique to measure conformational changes in the probe molecule at the nanometer scale."

Comments

dependent upon DNA hybridization

[Bowling Moses]

I just skimmed the article late this evening (early this morning? Whatever.). Anyway, it looked like what they'd done was to attach a single-stranded DNA sequence at one end to a slide, the other end is attached to a 1-micrometer diameter bead. Charge repulsion between the bead and the slide stretches the DNA strand, keeping it under tension. DNA with various sequences then can be introduced into the system, if they match the opposite strand of the fixed DNA strand well, then it will hybridize forming a double stranded DNA. Double stranded DNA forms a double helix structure which is more "fixed" structure than single stranded DNA, which can range from nearly linear to a random coil depending in part in the amount of tension its under and the sequence. Regardless, if there is a hit then the distance between the bead and the slide will change as the DNA is hybridized into a double strand, forming the double helix that we've all seen in biology textbooks. One problem is that multiple different DNA strands can hybridize nearly as tightly as an exact match, for example if we have the sequence 5' ACTGACTGACTG 3' then 5' CAGTCAGTCAGT 3' will bing to it, but so will 5' CAGTCAATCAGT 3', which differs by only one position. I hope I did that right, it's late, but anyway you can still get hybridization of DNA molecules that are only very similar but are not quite identical. This study used DNA strands 10's of nucleotides long so being off by one or even a couple of positions will still result in tight binding, although this can be tweaked a bit by messing with the DNA concentration; lower concentrations will favor more exact matches in general. But still, cool idea.

Re:dependent upon DNA hybridization

[sam_handelman]

Lower concentrations will also favor no match whatsoever, in general.

What matters is the relative concentration of the target DNA sequence, and of all other remotely similar sequences. Roughly speaking, for each base pair that is different between your ideal target and your actual target, you get a difference of a few kcal/mole in binding energy.

Every 1.36 kcal/mole (roughly) corresponds to a ten-fold decrease in binding affinity.

So, roughly speaking, a single-nucleotide mis-match is going to have 1/1000 times the binding affinity of a perfect match to the probe. This means, that under IDEAL circumstances, you can detect your target against a background of 1000-times its own concentration in single-base substitutions. Of course, under circumstances where your probe is long enough that random DNA will tend to bind indiscriminately, this won't work.

Contamination with single-stranded binding proteins, which do exist, might also be a confounding factor, either giving you a false positive or fouling up your probe.

Anyway, this may or may not be good enough for any particular application. I suspect that this technique will never actually be as sensitive as PCR, wherein the binding-affinity experiment is effectively "repeated" each replication cycle. If you choose a sequence carefully enough - and use a longer probe so that close matches are not so likely to appear at random (a ten nucleotide probe appears one in every 2^20 ~= 1/billion times, at random. The human genome is likely to include one instance of every decamer,) you might get performance good enough for the applications they describe.

Re:dependent upon DNA hybridization

[Sgt York]

I suspect that this technique will never actually be as sensitive as PCR

I disagree. While PCR can detect single copies of sequences in theory, it rarely (if ever) does so in common practice. There are tweaks you can do that will help amplify the system, and use of specialized detection equipment can get it down pretty low, but in my experience, anything less than ~100 copies of target is unreliably detectable, at best.

As for targets, a decamer is good in theory, but in practice you run into several problems using a probe that small. Most people use 20mers for PCR and sequencing applications.

This technique is a vast improvement over current technology for nucleotide detection. I've spread the paper around the department here, and people are talking about it quite a bit. It is in no way ready for prime time, and the brief article supplied by the link paints a rosy picture. The paper is more informative, and clues you in that due to steric influences, this can't be applied to any sequence bigger than (probably) about 100bp. It's still useful in its current form, but you'd have to abuse your sample before you could use it. It's primary use has not been lost on our gene regulation people. You can use this to easily see if protein x binds to sequence y and deforms it. THAT is new...and can really help some people around here.