New Method To Revolutionize DNA Sequencing

An anonymous reader writes "A new method of DNA sequencing published this week in Science identifies incorporation of single bases by fluorescence. This has been shown to increase read lengths from 20 bases (454 sequencing) to >4000 bases, with a 99.3% accuracy. Single molecule reading can reduce costs and increase the rate at which reads can be performed. 'So far, the team has built a chip housing 3000 ZMWs [waveguides], which the company hopes will hit the market in 2010. By 2013, it aims to squeeze a million ZMWs [waveguides] onto a single chip and observe DNA being assembled in each simultaneously. Company founder Stephen Turner estimates that such a chip would be able to sequence an entire human genome in under half an hour to 99.999 per cent accuracy for under $1000.'"

Comments

Gattica...

[Anonymous Coward]

Gattica, here we come!

But at least there's Uma Thurman.

Re:Gattica...

[Anonymous Coward]

Lets see your genetics save you from my Hanzo sword.

Re:Gattica...

[CorporateSuit]

but...

We have Uma Thurman NOW...

Re:Gattica...

[chill]

Unless we perfect cloning, there are going to be a few billion people who find little consolation in that.

Re:Gattica...

[Anonymous Coward]

Unless we perfect gene therapy, there are going to be a few billion people who find little consolation in that.

Fixed that for you. Cloning wouldn't help the people who are alive now.

Re:Gattica...

[Anonymous Coward]

Sorry... but, "Gattica"?! Did you completely fail to understand the meaning of the name of the movie? (since when was there an "i" nucleotide?)

Re:Gattica...

[KeithJM]

Unless we perfect cloning, there are going to be a few billion people who find little consolation in that.

Fixed that for you. Cloning wouldn't help the people who are alive now.

The clones would be alive, and imperfect because of the errors in cloning.

Re:Gattica...

[Noddegamra]

Since inosine.

Re:Gattica...

[treeves]

OK, he should have asked more specifically, "since when is there an 'I' nucleotide that belongs in DNA?"

99.3% accurate?

[Valdrax]

That's, what, 28 incorrect base pairs out of 4000? I'm not a biologist, but is this considered an acceptable error rate? Even the hopes of 99.999% accuracy seems really awful when there are about 3 billion base pairs in a human genome.

I realize that we aren't going to be trying to make a cloned copy from this data, but what uses is this "good enough" for?

Re:99.3% accurate?

[imamac]

I realize that we aren't going to be trying to make a cloned copy from this data...

What makes you so sure? Who knows where this will lead?

Re:99.3% accurate?

[the_humeister]

It's not too bad. I don't think the human version of the polymerase has a better error rate. However, while being in a biological entity, DNA replication also has other integrity checks.

Re:99.3% accurate?

[Anonymous Coward]

Coverage

Re:99.3% accurate?

[aoeusnth]

That's, what, 28 incorrect base pairs out of 4000? I'm not a biologist, but is this considered an acceptable error rate? Even the hopes of 99.999% accuracy seems really awful when there are about 3 billion base pairs in a human genome.

I realize that we aren't going to be trying to make a cloned copy from this data, but what uses is this "good enough" for?

More than good enough for forensic work at least, I'd wager.

Re:99.3% accurate?

[Maximum Prophet]

How many errors are introduced during normal human reproduction? The dogs they've cloned so far are less than 99.999% identical.

Re:99.3% accurate?

[Anonymous Coward]

It's common practice in bioinformatics to measure the same data repetitively in an effort to reduce the error. While 0.993 isn't very good, (0.993)^3 is pretty awsome. In practice, the errors might be correlated (as in a flaw in the measuring system), so the benefit of re-measuring might not be exponential...however it should be darn close.

Re:99.3% accurate?

[jbeaupre]

Given the expense of doing an entire genome, alternative is a 25% accuracy rate. What 99.3% does is let you do a bulk scan looking for interesting areas. Prospecting. Now you can adjust therapies the match likely genome sequences. "Ah Ms. X, I see you likely have gene XYZ. Medication A, B, and C won't work for you so let's try D."

In other words, you don't need perfect results to now bias the odds in you favor.

Re:99.3% accurate?

[ccguy]

Well, depends if those 28/4000 errors are the same in each run or not.

If they can sequence the whole thing in less than 30 minutes one time with a 0.001% "read" error rate, my guess is that they can get it probabilistically near 100% correct in 2 hours or so.

By the way, what's the current error rate? Is it 0? (just asking)

Re:99.3% accurate?

[morgan_greywolf]

1 Hour Genome Sequencing: 30,000 errors or less or YOUR MONEY BACK!

Re:99.3% accurate?

[nbauman]

I want a cute little baby velociraptor!

Re:99.3% accurate?

[mapkinase]

You can decrease the error rate by increasing coverage from 15-fold to N-fold

Re:99.3% accurate?

[Stile 65]

Do it several times over with different cells and "vote" on the inconsistencies between trials. If 5 out of 7 copies of the DNA look like the base at position X is tyrosine, then it's most likely that it's tyrosine.

Re:99.3% accurate?

[msh104]

I suppose that running it twice or trice will increase the accuracy a lot more.
Which makes it still blazing fast.

Re:99.3% accurate?

[cobaltnova]

3 billion * 99.999% = 30. That's not "really awful." That's pretty darn good! If you need better results, sequence the same DNA 3 times in an hour and a half to get upwards of ten 9's accuracy (assuming the errors are uncorrelated, of course).

Re:99.3% accurate?

[Hatta]

When I have sequences done the conventional way, I get less than 1000 base pair reads back. Generally 2 or 3 are ambiguous enough that the machine reads them incorrectly or not at all. 28 out of 4000 is the same as 7 out of 1000, so this is roughly the same magnitude of error. Less accurate than what we use now, but more economical to do really large sequences.

I don't know how the method works (site is slashdotted anyone got a DOI for the paper?) so it's hard to tell whether repeating the reads would get you a different set of random errors or just give you the same errors over again. But I'd imagine that if you really needed extremely high accuracy, you could just have it done several times and pick the consensus sequence.

Re:99.3% accurate?

[Anonymous Coward]

Re: mistakes and inaccuracies...

You run two or three trials and do "a check sum" ...a la Raid inter leafing...errors stand out and are discarded..

Re:99.3% accurate?

[bigattichouse]

I want mammoth burgers, the original human food. oh, and organized mammoth hunts on the canadian tundra.

Re:99.3% accurate?

[philspear]

That's, what, 28 incorrect base pairs out of 4000? I'm not a biologist, but is this considered an acceptable error rate? Even the hopes of 99.999% accuracy seems really awful when there are about 3 billion base pairs in a human genome.

That's a very good question, but consider that 100% is impossible. Even the cell's own machinery, under development for millions of years, makes mistakes at a frequency that would be lethal if that's all there was.

In this case, the error rate seems in the neighborhood of rival techologies. The way to deal with it is the same way the cell uses: redundancy. Sequence segments or the whole thing more than once, the likelyhood of bases in error is significantly decreased. If you run 3 sequencings, there's even a smaller chance that you'll have an error 2 out of 3 times.

Anyway, thanks to the way the genome works, the vast VAST majority of errors won't matter as much: non-coding DNA, introns, the codons themselves have a high degree of redundancy (IE TCT codes for serine, if the last T gets read as a C accidentally, it will still give you serine.)

Granted, there are important uses the sequence itself has. Fortunately most of those themselves can be done redundantly. If you're trying to run an in-situ hybridization for a sequence and your probe has one or two errors in it, my understanding is that would mess up the in-situ, but people often use more than one probe.

I should point out I'm not a biochemist, so take everything with a grain of salt. As with everything, I am definitely not error free.

Re:99.3% accurate?

[evilNomad]

If you got B on the second run you'd be pretty sure it was incorrect.. ;-)

Re:99.3% accurate?

[Rayban]

Sorry, you're asking for the impossible - I've never seen a well-organized mammoth hunt.

Re:99.3% accurate?

[scorp1us]

There is a saying from the old sailing days. "Never set sail with two compasses". One is ok, three is better. But never two. The paralysis from not knowing which is right is far worse than being wrong and correcting later.

Re:99.3% accurate?

[0prime]

Besides the current near impossibility of building a clone instead of using existing DNA, he made that statement because the method for sequencing is only 99.3% accurate. So basically, you completely missed his point, since he was saying he understood they may not need 100% accuracy (not cloning), but wondered where something with a 99.3% sequencing accuracy would be applicable.

Re:99.3% accurate?

[philspear]

... and to prove that last point I just realized that I was redundant with the non-coding DNA and introns. I think. No wait, I meant to do that, this way if you misread "introns" it will still be covered by the "non-coding DNA" bit. And that's the last biochemistry joke out of me today.

Re:99.3% accurate?

[thesandtiger]

There's an easy and obvious way around this - just run 3 simultaneous instances and error-check by consensus. Still able to run the whole thing in under a half hour and still pretty cheap at ~$3000.

Re:99.3% accurate?

[imamac]

Oh well. At least it's not the first time somebody missed a point on /.

Re:99.3% accurate?

[Tubal-Cain]

At the very least, this method can be a cheap way to acquit suspects. Those that come up positive can ask for the more accurate test.

Re:99.3% accurate?

[untermensch]

The idea is to sequence each portion of the genome many times over. With enough redundancy you can detect these errors, so your final annotated copy would have a much higher accuracy.

Re:99.3% accurate?

[Mordstrom]

Perhaps the Borean Tundra...

Re:99.3% accurate?

[shaitand]

If you were sequencing DNA and got a B then you'd seriously need to recheck the equipment (or the competence of the operator). Perhaps a T or a G, or even a C but never a B.

Re:99.3% accurate?

[prograde]

It's common practice in bioinformatics to measure the same data repetitively in an effort to reduce the error.

It's common practice on Slashdot to read the article before posting. From the abstract of the Science article:

Consensus sequences were generated from the single-molecule reads at 15-fold coverage, showing a median accuracy of 99.3%, with no systematic error beyond fluorophore-dependent error rates.

So that's 99.3% after averaging 15 reads. Not exactly replicating the same read 15 times..more like taking random starting points and aligning the results where they overlap, so that each base is covered in 15 different reads.

Don't get me wrong - this is really cool, and a massive speed-up over current "next-gen" sequencing. And I'm sure that it will get better.

To answer the GP - yes, this is an acceptable error rate, for now.

Re:99.3% accurate?

[fuzzyfuzzyfungus]

I suspect, given that this system is fast, massively parallel, and not terribly accurate, that (should it reach production use) its users will end up relying on multiple runs and clever error correction techniques. As long as the .7% errors are random rather than systematic, you should be able to reduce the effective error rate by using "best x out of y", with X and Y chosen for your budget and risk tolerance.

It probably also will find a niche, in spite of its error rate; because there is just so damn much DNA that biologists would like to sequence. Millions of known species, loads more unknown, even just grabbing a sample of the environment and sequencing whatever is inside which can turn up loads of stuff we didn't even know was there. I strongly suspect that there are large numbers of situations where an approximation that we can afford is far more valuable than a fully accurate sequence that we can't.

Re:99.3% accurate?

[fracai]

Oh well. At least it's not the first time somebody missed a point on /.

Don't you mean "Oh well. At least it's not the first time somebody missed a point on /".

Re:99.3% accurate?

[bennomatic]

Actually, you're off by a factor of 1000.

3,000,000,000 * 0.99999 = 2,999,970,000

So your error count is 30,000.

Re:99.3% accurate?

[Homericus718]

Although this repetition you speak of is important, it is far less beneficial than exponential. (0.993)^3 is actually less accurate (0.979), I would guess you meant to do (1-(1-0.993)^3) = 0.9999997. Also, re-measuring data does not give an exponential error decrease. Most signal to noise levels go up as the square root of the number of samples taken. This is assuming the error isn't systematic.

Re:99.3% accurate?

[cat_jesus]

Do the sequencing three times. So you spend 1.5 hours instead of .5 hours. If you're really worried do it 7 times. That should be enough to weed out the errors.

Re:99.3% accurate?

[Anonymous Coward]

Come on moderators, whip out your calculators and check. (0.993)^3 is even worse than 0.993 in terms of accuracy.

Re:99.3% accurate?

[marcosdumay]

So, instead of $1000, you must spend $3000 and get the same result in one hour that we needed 10 years and several millions to get at the 90's... I see, that is useless.

Re:99.3% accurate?

[peter303]

One in 10E8 is the DNA base-pair copy error rate. Even so thats around 60 when a sperm meets egg. Another much more when there a trillion somatic cells dividing on average 50 times each in a human lifetime. The vast majority are errors are neutral, but accumulating ten or so specifically unluckly ones in a cell may be a cancer.

Re:99.3% accurate?

[MyLongNickName]

This assumes that the method simply has a random chance of getting each data point wrong. What if it is something systematic with the method that causes it to read one gene wrong? In other words, it reads the gene as a 'T' every time despite it really being an 'A'. No matter how many tests you run, it will still result in a wrong answer.

Re:99.3% accurate?

[timeOday]

While 0.993 isn't very good, (0.993)^3 is pretty awsome.

No, 0.993^3 is only 97.9%; how about 1-(1-0.993)^3 :)

Re:99.3% accurate?

[Adriax]

Or you could run a parallel processing setup, 3-5 sequencing chips all given the same sample at the same time. More expensive, but you'd get that effective 100% rate in the half hour time.

$5k for a genetic sequencer that could give effectively 100% accuracy in half an hour would be pittance for pretty much every hospital in the US.
Hell, the first malpractice lawsuit it prevents (detect a disorder that would make a commonly used treatment crippling or fatal to the patient) would pay for the machine 1000 times over.

Re:99.3% accurate?

[smoker2]

Thrice.

Re:99.3% accurate?

[ogdenk]

Could also be useful the other way around.

1.) Rapidly sequence a suspect's DNA

2.) Find a cheap way to make "good enough" copies

3.) Plant evidence

4.) ????????

5.) PROFIT!!!!

Re:99.3% accurate?

[Anonymous Coward]

It's super fine! It's very very good. I think there is some confusion since they are using different metrics for accuracy. The 99.3% accuracy likely does mean an expected 28 errors per read of 4000 base pairs. But their 99.999% accuracy likely means that there's 99.999% confidence that there are no errors at all. Here's a crash course in DNA sequencing with hand waving and generalization. You slice the DNA into (mostly) random bits. With old methods they were about 20 BP long. With newer 454 pyrosequencing I think they were 100 BP long, or longer. This new technique uses about 4000 long fragments. Then you use some kind of magic to look at those slices. Pyrosequencing gets its name because you have reagents involving pyrophosphate, that attach to your base pairs. So one at a time you can add reagents to combine with the chain, and when the pyrophospate is released it glows. The brighter, the more of the same BP in a row there was. So you put them all on a slide and can do thousands at once. Mix reagents, your CCD looks for glows. Neato! Haven't read this article to see how it works but it doesn't really matter how, so long as it does work.

But, what do you do with these nice little fragments? (not so little if they're doing 4000 long reads!) Well, you run a pattern matching algorithm to line them up! You are right you'll have 28 errors on average per fragment. But, with 4000 long fragments, if you have overlap of say 1000, even with all 56 errors occurring in that overlap (bad luck) I think you'll find that the odds of this being the correct matching, with 56 errors, are massively massively higher than the odds of the other 944 base pairs matching by random chance or read errors! So these errors won't disrupt the assembly, at least it would be amazing if they actually did. Now typically one would use 15 fold overlap. That is, you run enough fragments so that, statistically speaking, you have high confidence that every basepair will be a member of at least 15 different fragments. So, after assembling, even with some assembly errors, you can do majority voting. If at least 9 out of those 15 (you could in fact use 8 but I'm leaving 1 extra!) all agree, then that's the base pair that goes there!

Now that that's "explained," what are the actual odds of an error in a genome of length 3*10^9? Well, lets imagine we have the whole human genome done in this way, with reads of length 4000, and with 99.3% accuracy for any given base pair. Well, then in our assembled genome, the odds of an error at a given basepair will be (0.993)^14 * (0.007), odds of two, thee, also calculable. However, when assembling, because the reads are so long, you can be extremely confident that that is where they go, so one error won't disrupt you at all. Even 6 is fine, because the other 9 fragments that overlapped that BP all agree still. The odds of 7 or more errors on the same basepair is something like 10^(-15). The probability of having this occur ZERO times out of 3 billion base pairs is 99.9997%. So that's where their number comes from. It's not that out of these 3 billion base pairs, 99.9997% are right, meaning an expected OVER 9000 base pairs will be wrong. It's saying that with 99.9997% confidence, there are ZERO errors. If there is an error, there's again many 9s chance that it's only a single error. That's pretty damn good! Nearly a 1 in a million chance of even a single error! Of course, the errors won't be uniform, and it may be position dependent (pyrosequencing sure is, I don't know about this technique though). And the coverage won't be uniform, either, so a BP that already has a higher than average proportion of errors could, by bad luck, end up with a low amount of coverage to boot! But I'm sure if you double your coverage for an extra $1000 you can be super sure of the coverage, and can also afford to be extra pessimistic in your error rates, and still get your 99.9999% chance of an error-free sequence.

Re:99.3% accurate?

[Anonymous Coward]

from the science article:

"Consensus sequences were generated from the single-molecule reads at 15-fold coverage, showing a median accuracy of 99.3%, with no systematic error beyond fluorophore-dependent error rates."

The method basically relies on not getting an accidental fluorophore to float into the reading frame, which happens about .7% of the time. This is a random event and non-systematic. Taking twice as long being exponentially more accurate

Re:99.3% accurate?

[Anonymous Coward]

Run it a few times over the same sequence, and then take the most popular results. If the errors are generated randomly(I don't actually have any reason at all to believe this is the case), rather than something systematic in the process, the chances of a collision of false reads is pretty small. If it only takes 1/2 an hour to run, so if it was required more samples wouldn't cost that much(compared to something like artificially creating an organism).

It will probably be good enough for forensic work and most biological stuff. Our current methods are pretty far from perfect.

Re:99.3% accurate?

[jd]

That's because Mammoths have no opposable thumbs and therefore no means of becoming organized.

Re:99.3% accurate?

[wealthychef]

You can repeat the analysis several times to increase accuracy, I would think. Do the analysis 3 times and it might increase to 99.9999997%, for example.

Re:99.3% accurate?

[wealthychef]

The point is, which pieces are incorrect? It's highly likely to be slightly incorrect, but you want it to be highly likely to be completely correct.

Re:99.3% accurate?

[przemekklosowski]

While 0.993 isn't very good, (0.993)^3 is pretty awsome.

(0.993)^3 would suck actually, resulting in 0.979---but fortunately the error rate is 1-(1-0.993)^3), i.e. a pretty awesome .9999997 (assuming independent errors and such)

Re:99.3% accurate?

[Anonymous Coward]

By the way, what's the current error rate? Is it 0? (just asking)

No: the error rate depends on the technology. First of all, in all sequencing technology, the error rate increases as the sequence gets longer. That is, the probability of an error in the first bases is much smaller than a probability of an error in the latter bases.

"Sanger sequencing", the most established method for sequencing, tends to peter out by the time it gets to about 1000bp reads. However, Sanger sequencing is slow and expensive.

Newer high throughput sequencing technologies (Illumina, Applied Biosystems) can generate much more sequence at once, but compromise in that read length is about 25-100bp, and the sequences are more prone to errors (again, especially in the latter bases). The Applied Biosystems technique has a built in error checking code, which is pretty trick.

Finally, there is a medium range solution (454) which gives ~400bp reads, at a lower throughput.

So, taking that all together, the point is that none of these are good enough to just sequence the human genome in a single shot. Instead, the standard approach is to sequence at a certain "coverage". Practically, sequencing a genome de-novo with Sanger sequencing often takes about 10x coverage (e.g. need to sequence the human genome about 10 times over) to get confidence in the final sequence. Using more high throughput sequencing methods, we need more like 20-40x coverage to get an acceptable error rate. This also has to do with assembling the reads together, but I won't get into that. Lets just say, the human genome is not a random set of bases, and has much repetitive sequence.

So, take home, no matter how you sequence, getting as high-accuracy pass of the human genome, you can generally make up quality with quantity. For the human genome, 10X coverage is 30B bases.

And, lastly, there is the effort of "resequencing", where you have a reference genome, and are simply looking for variants. For example, we are not going to sequence many people de-novo, but instead in comparison to existing sequencing to identify difference. Here, lower coverage is acceptable.

Re:99.3% accurate?

[Anonymous Coward]

If you RTFP (requires subscription), no systematic errors were detected
http://www.sciencemag.org/cgi/content/full/323/5910/133

Re:99.3% accurate?

[m93]

I realize that we aren't going to be trying to make a cloned copy from this data, but what uses is this "good enough" for?

It's most likely good enough to deny you health coverage. Pre-existing condition? Now risk can be assessed on pre-existing genes.

Re:99.3% accurate?

[Red Flayer]

It's common practice on Slashdot to read the article before posting.

In what alternate universe are you reading slashdot?

prograde (1425683)

Oh. You'll figure it out, given time.

Re:99.3% accurate?

[noidentity]

I realize that we aren't going to be trying to make a cloned copy from this data...

What makes you so sure? Who knows where this will lead?

Probably to people thinking a single live mirror of themselves is sufficient, until they get sick and find the mirror is also sick. I actually think I read a similar story recently...

Re:99.3% accurate?

[Anonymous Coward]

The biological mechanisms of reproducing DNA have an embedded error not far from this rate.

One could argue that 5-nines accuracy at time t is just as good as 100% accuracy at t+2 (after the DNA has replicated a couple of times).

Re:99.3% accurate?

[Red Flayer]

0.993)^3 would suck actually, resulting in 0.979---but fortunately the error rate is 1-(1-0.993)^3), i.e. a pretty awesome .9999997 (assuming independent errors and such)

No, your formula results in an accuracy of .999, not .9999997. Sure, it's better than .993, but nowhere near as accurate as you wish.

Did you assume that .993 was accurate to .9930000?

Furthermore, this entire formulaic analysis is useless without considering statistical error margins.

Re:99.3% accurate?

[Arancaytar]

The final estimate is 99.999%, presumably including some redundancy or error checking.

This is still one error in 100,000 base pairs, which is tens of thousands of errors in the 3 billion pairs of the human genome.

Of course, we're not talking about computer code here - this code has withstood (or rather adapted to) millions of years of inaccurate copying. Most of the errors would probably hit the vast majority of inactive areas, or be compensated by the natural redundancy mechanisms.

Re:99.3% accurate?

[Neeperando]

This is actually how all DNA sequencing is already done. Most genome projects try to get 8-10X coverage, which means they sequence 8-10 times as much DNA as exists in the genome in the hopes that most bases occur in 8-10 of the reads that make up each location in the final sequence. Typically, if all 7 of 8 are the same, the 8th is considered an error, whereas if 5 of 8 are A and 3 of 8 are T it is treated as a SNP (single nucleotide polymorphism. In layman's terms, a difference between your mom's version and your dad's version of the same sequence).

It seems they already figured this into their error rate:

We have shown that with just 15 molecules, a consensus sequence with 99.3% median accuracy can be formed with no detectable sequence context bias and a uniform error profile within reads.

They go on to say this may prevent it from being useful for sequencing new genomes from scratch, but it will may help with the kind of applications currently being used in the medical community, like identifying genes that cause cancer, etc.

Re:99.3% accurate?

[Hordeking]

Sorry, you're asking for the impossible - I've never seen a well-organized mammoth hunt.

At least not one organized by Canadians...

Re:99.3% accurate?

[Arancaytar]

Don't know - but it doesn't need to be 0 to be reliable. The genome is quite resistant to random mutation, having been subjected to it for all its existence.

Re:99.3% accurate?

[Metasquares]

You'll figure it out, given time.

You must be new here :)

Re:99.3% accurate?

[TooMuchToDo]

Run the data among multiple chips for verification. Run amongst enough chips, the errors should be detectable/correctable.

Re:99.3% accurate?

[Anonymous Coward]

I got a B in DNA sequencing. My parents STILL grounded me.

Re:99.3% accurate?

[TooMuchToDo]

Unless of course all three of your compasses are giving you different readings. In that case, you simply yell "Where the hell is my sextant."

Re:99.3% accurate?

[amalyn]

Systematic errors could be identified by correlating results with other DNA sequencing results.

Using a large sample, like the proposed Personal Genome Project [unsure if they have gotten in touch with any of those who expressed interest in participating] could be useful in showing any systematic mis-reads, as long as the Personal Genome Project is using another method to sequence the participant's DNA.

Re:99.3% accurate?

[sorak]

It's common practice in bioinformatics to measure the same data repetitively in an effort to reduce the error. While 0.993 isn't very good, (0.993)^3 is pretty awsome. In practice, the errors might be correlated (as in a flaw in the measuring system), so the benefit of re-measuring might not be exponential...however it should be darn close.

I don't mean to trample your point, but three iterations wouldn't give a (0.993)^3, (which would equal 97.9%). The odds of an error would be 1-(0.007)^3, which would actually be 99.9999657%.

Re:99.3% accurate?

[Rasperin]

Clone a human, get a large singing dinosaur! Close enough.

Re:99.3% accurate?

[Anonymous Coward]

I've had mammoth burgers and they tasted a lot like your mom; old and hairy.

Re:99.3% accurate?

[Vexorg_q]

I am a biologist, and I can tell you that's about the same error rate as other sequencing methods. But, like all things, there are a lot of variables to consider (such as inital sample quality and source) And for any research, sequencing should be repeated and compared (which is done).

Re:99.3% accurate?

[Crookdotter]

The errors won't be in the same place every time, so you'd run it 5 times and take the majority base seen. Close to 100% accuracy with only 5 simultaneous runs over 30 mins and $5k? Sounds like a winner.

Re:99.3% accurate?

[DNS-and-BIND]

Humans don't reproduce by cloning, dumbass. Or maybe that's the only way you can think of to reproduce, the rest of us have other ideas...

Re:99.3% accurate?

[Kazrael]

(.993)^3 is less than .993
You're taking less than a whole times less than a whole times less than a whole. .993^3 = .979146657
Perhaps what formula you meant was (1 - ((1 - .993)^3) = 0.999999657

Re:99.3% accurate?

[Kazrael]

Effing matching parenthesis.
(1 - ((1 - .993)^3)) = 0.999999657

Re:99.3% accurate?

[aled]

That's, what, 28 incorrect base pairs out of 4000? I'm not a biologist, but is this considered an acceptable error rate?... but what uses is this "good enough" for?

three-eyed fish?

Re:99.3% accurate?

[Nefarious Wheel]

I want a wallet made out of Borean Leather, +50 Armor.

Re:99.3% accurate?

[aliquis]

Little I hope, the girl who wants to mate with me has to be really retarded and I don't want that to affect our kids :D

Re:99.3% accurate?

[Anonymous Coward]

> 1.) Rapidly sequence a suspect's DNA > 2.) Find a cheap way to make "good enough" copies > 3.) Plant evidence Huh?! Step one only works if you already have a suspects DNA, and if you do, you can plant it. Why use millions of dollars of equipment to clone a suspect? Just plant the cells you already have, or (for the mad scientists only) encourage the cells you have to grow a bit so you have more. No need EVAR to sequence a suspect's DNA if you want to plant it.

Re:99.3% accurate?

[Thiez]

Posting to undo moderation. Modded parent Overrated by accident.

Re:99.3% accurate?

[aliquis]

lol, you got moderated 5 informative but your math is flawed.

Do it three times to get 97.9% accuracy instead of 99.3? Great! :D

Guess it should be something like:
1 - ((1 - 0.993)^3) = 0.999999657 ? =P
Actually I have no idea, too tired to think about it but not 0.993^3 atleast :)

Re:99.3% accurate?

[Nefarious Wheel]

Unless of course all three of your compasses are giving you different readings. In that case, you simply yell "Where the hell is my sextant."

Astrolabe FTW!

Re:99.3% accurate?

[ogdenk]

Dude.....it's a joke......really.

Re:99.3% accurate?

[Anonymous Coward]

In a trailer park with massive amounts of beer? Quite a bit it seems.

Re:99.3% accurate?

[Fluffeh]

While finding these error rates and posts like the parent are cool, what made me really LAUGH in this article wasn't how many errors or lack of errors or the timing, but that even working out to 2013, they knew exactly how much they were going to charge for this.

Talk about medicine for healing rather than profiteering eh?

Re:99.3% accurate?

[Fluffeh]

I must have missed that quest...

Re:99.3% accurate?

[Anonymous Coward]

the whole genome is not sequenced in one linear read. multiple random fragments are sequenced. each fragment has overlap with numerous other fragments. this redundancy ensures that the accuracy is very high after the final assembly.

Re:99.3% accurate?

[AmigaMMC]

99.3% accuracy might not seem much when looking at 3 billion pairs but remember that a lot of those pairs are dormant or useless (AFAWK).

Re:99.3% accurate?

[Anonymous Coward]

Human DNA polymerase error rates are usually about 1x10^-5 (or -6) when they include the proofreading activity (which they obviously do in vivo). Recombinant versions sometimes don't have the proofreading activity because speed is sometimes valued over accuracy.

1x10^-5 = 99.999%

Re:99.3% accurate?

[Anonymous Coward]

Nobody sails on the poles, dude.

Re:99.3% accurate?

[WiFiBro]

What's this cloning fascination all over this page? If you want to clone, why would you need to first read out the entire genome? Nature is doing pretty well in copying DNA if you only find a nice cell to start with.

Further, in DNA-matching suspects they are doing pretty fine now with a handful of typical pieces of DNA. Don't need the whole genome for the first step.

Re:99.3% accurate?

[dbitch]

I realize that we aren't going to be trying to make a cloned copy from this data, but what uses is this "good enough" for?

It's good enough to make Ivory: a pure 99.44%.

Re:99.3% accurate?

[dublin]

Good enough for Gattaca, maybe? After all, how accurate do you have to be if the only fallout is denying some poor unfortunate soul his rights as a human being?

I'm *really* starting to think maybe Bill Joy was right, and that these technologies are far, far more dangerous than we currently realize. The hearts of men are twisted and dark - technology itself is neither good nor evil, but can be a very effective amplifier of our fallen nature. (Look no further than current events in Iran...)

Re:99.3% accurate?

[Anonymous Coward]

Yes, thats acceptable. Current methods of sequencing require that segments of different lengths be sequenced multiple times and the results of the individual sequencings are compiled to create a more accurate (correctly mapped) result.
-AC

Re:99.3% accurate?

[icejai]

I wonder if this is the evolutionary benefit to having so much 'junk' dna; to absorb copy errors.

Re:99.3% accurate?

[GreatDrok]

All sequencing methods have error rates and this is very low given the read lengths they are looking at. Every method currently used goes for multiple coverage where the same piece of sequence is run multiple times. Read errors will be somewhat randomly (I say somewhat because some techniques produce more errors later in the read) so by reading the same sequence multiple times you can exclude the errors and build a 100% correct sequence. The biggest problem with current next gen sequencing techniques is that they have short reads (Illumina/Solexa are about 36bp) and this makes sequence assembly a challenge. You need high levels of coverage to detect errors versus SNPs (real mutations) and will also have problems traversing repetitive regions of sequence. 4000bp reads, even with errors, will help enormously and reduce the complexity of de-novo assembly (assembling without a template to map against) so we will be able to sequence many more species accurately and quickly.

Re:99.3% accurate?

[Rich0]

Are you suggesting these guys should just fund their own lab for five years (employing who knows how many people) without a care for whether it will ever make money?

Sure, the government can afford to not worry about profit, but most other people who spend five years on something expect to get something out of it (like their costs for starters).

Unless you're of the mindset that all non-trivial research of any kind anywhere should be funded by taxpayers profitability is going to be a concern. And it isn't like making a profit costs the patients more - they're paying for the R&D one way or another (just via taxes instead of fees).

And who is to say that these guys even care about the money in the first place? Aiming to hit a particular price point is a goal in both the private and public sectors (why else would scientists dream about space elevators but for the possibility of making $1000 probles and launching one a week?).

Re:99.3% accurate?

[Anonymous Coward]

uhhhhh... go try putting 0.993^3 in your calculator. I think you mean (1 - 0.993)^3 and then subtract that from 1.0

Re:99.3% accurate?

[seasunset]

It depends on what you want to do. Note that existing methods also have problems of this kind.

I remember listening to a world wide specialist on the issue (with papers and software published on handling errors in genetic datasets) where he talked about error rates of 10% in some cases (mostly human introduced: like reading protein gels wrongly or just plain typos on spreadsheets).

I have tested myself the HapMap project (sequencing of human SNPs in several populations) and the error rate, while very low, doesn't allow for studies of mutations from parents to offspring (the noise - error - is orders of magnitude bigger than the signal - very low mutation rates from parent to offspring).

I would say 28 out of 4000 is quite good. Although it would be great if it was random (like you could repeat the experiment and get a different set of errors). This would allow to go lower (at the expense of more experiences).

Re:99.3% accurate?

[Richard W.M. Jones]

There is a saying from the old sailing days. "Never set sail with two compasses"

Is there?

Rich.

Re:99.3% accurate?

[Escogido]

http://velociraptorz.org/

You're welcome.

Re:99.3% accurate?

[Anonymous Coward]

The Applied Biosystems technique has a built in error checking code, which is pretty trick.

It is only error checking if in there are NO ERRORS in the error correction code. Otherwise you also have to account for the errors in the encoding (which on our machine is >10% error).

Re:99.3% accurate?

[tobias.sargeant]

Actually, B in the IUPAC code corresponds to [CGT]. It's unlikely (more likely that the base would be called as N=[ACGT]), but possible, that that would arise during a sequencing run.

Re:99.3% accurate?

[g_assembly]

Read lengths of 4000 with 28 miscalled bases would be awesome, totally awesome. Current "next-gen" sequencers typically generate 100x coverage over the entire genome so, to use a puzzle analogy, it's like having enough puzzle pieces to build 100 complete puzzles. This coverage lets you find and correct sequencing errors to generate a "consensus sequence." So if you can successfully identify 100 code fragments covering the same piece of the genome and 8 have an "A", "G", or "C" in position 11 but 92 have a "T" in position 11, you can be reasonably confident that the consensus call of "T" is the correct nucleotide base and the other calls are sequencing error. The read length of 4000 as opposed to say, 25, lets you span areas of tandem repeats, found in certain organisms, that confound assembly algorithms and give you bad assemblies. Hope this helps.

Re:99.3% accurate?

[Anonymous Coward]

Right...and, if the first base is wrong, the whole sequence is a toss, right? Still, cool idea if they can get it to work better.

Re:99.3% accurate?

[fatdefacto]

DNA is sequenced using a method called shotgun sequencing, so essentially for every base pair of DNA there will x amount of reads covering it which are assembled together using a shortest common superstring algorithm. The acceptable rate of error is 1 in 10000 to publish a complete genome. Also 454 sequencing produces 250-500bp reads. The OP must be thinking of solexa or abi methods.

Re:99.3% accurate?

[Anonymous Coward]

No measurement in empirical science is meaningful without a corresponding measure of uncertainty.

Even seemingly simple discrete cases will contain the reality of error. Carefully annotated linguistic corpora and published materials consistently contain a small percentage of error. A large part of science is understanding the uses of "good enough". The perfect is the enemy of the good.

Re:99.3% accurate?

[Archangel Michael]

Me, personally, I'd rather see Carnivorous Mammoths hunting people ... but that is just me.

Re:99.3% accurate?

[jamstar7]

That means I'll finally be able to get a purple dino hide luggage set, including briefcase.

Looks pretty win-win to me...

Re:99.3% accurate?

[Clanked]

He was making a DNA joke that went right over your head. See: Base Pairs

Re:99.3% accurate?

[Anonymous Coward]

Ummm, 0.993^3 => 0.979

Re:99.3% accurate?

[davolfman]

You need at least 3 to have a good voting system (forgive me if I got the term wrong). That way when one part gives bad data the majority parts override it.

Re:99.3% accurate?

[Anonymous Coward]

worse comes to worse they re-run it, or sequence to the gene of interest

Re:99.3% accurate?

[Anonymous Coward]

Yes, expect this is the error rate after they sequence 20 times IIRC. It's all hype...

Re:99.3% accurate?

[Anonymous Coward]

It's an acceptable error rate for a proof on concept, for production sequencing it's appallingly bad. In order to be competitive with current next-gen players they need to increase throughput by an order or magnetite, decrease the error rate on a significant percentage of the bases to 1 in 1000 and increase the read length to 100bp+

Re:99.3% accurate?

[Anonymous Coward]

The sequencer is going to cost way more than 5k probably more like 250k. 5k would be the cost of the sequencing experiment. In any event this is a /very/ early proof of concept and in their calculations they are assuming they can get their error rate down at least 15 fold (they had to pile up 15 reads per base to get the error rate stated).

Re:99.3% accurate?

[Anonymous Coward]

1. They are already piling up 15 reads to get down to 28 out of 4000, which is pretty bad.

2. It's important that you can identify bases that are very likely to be accurate, and those that are not likely to be accurate. You can do this in Sanger and 2nd generation sequencing. Those bases that are wrong in your Sanger sequenced read, mostly likely have a poor quality score associated with them.

Re:99.3% accurate?

[cpricejones]

They picked a DNA polymerase that is hardy enough to not have those sorts of sequence biases. Or so they note in the article that previous researchers had characterized the DNA polymerase well enough to know that there are not many sequence biases.

Re:99.3% accurate?

[Vornzog]

BIG nit-pick. This is funny, until you realize that 'B' is a recognized character in the DNA alphabet. It is a degenerate base that means 'not A' (i.e. a mixture of C, G, T/U).

More degenerate codes here.

Sometimes these are artifacts of the sequencing technology. More commonly, there is actually an admixture of DNA in the starting sample. They are very common in RNA virus genomes, where the error-correction of the polymerase enzymes is crap compared to what mammals have evolved. This is a good thing for the virus, as it mutates quickly and escapes from immunity in the population.

In a bad sequencing run, you'll get a bunch of these mixed positions, and your data is crap. In an otherwise clean sequencing run, a mixed base position is extremely important information, and is often called incorrectly by both humans and automated base calling software, because people think that there must be exactly one non-degenerate 'right' answer.

This is actually a major problem as we start using '2nd generation' sequencing (454) and developing '3rd generation' sequencing like this article is describing. The amount of data being produced is moving past the point where humans can deal with it directly, so the assembly has to be automated. But the heuristics for tasks like calling bases are not yet up to the task, meaning that we sacrifice quality of information for quantity.

There are good opportunities in the field of bioinformatics for computer scientists who know enough biology!

Re:99.3% accurate?

[Vornzog]

By the way, what's the current error rate? Is it 0? (just asking)

Not 0. It depends on how many reads you have over the region. If you know the error rate and you have multiple reads over the same region, you can put a confidence score on your call. Most of the human genome stuff has something like 8-fold coverage, which means your error rate stays pretty low. For normal sequences.

Problems arise when you get into highly repetitive areas of the genome, or areas that can form secondary structure. These are often biologically interesting, but difficult to sequence, so the error rate is much higher.

The new sequencing techniques, like those discussed in this article, have a much harder time with single base repeats than traditional methods do. But they can sequence 'normal' DNA much more quickly.

I predict that you'll start seeing a lot more hybrid approaches. The easy stuff will get cranked out in bulk by 454 or the new solid-state technologies. You'll get a rough assembly, covering 98% of the stuff you want to know, and you can then go back and cherry pick interesting/concerning regions with more traditional methods (to knock down false positive mutations) and fill in gaps in repeat regions (useful for 'fingerprinting' DNA).

Re:99.3% accurate?

[Revek]

Don't ever expect perfection from first generation hardware.

Re:99.3% accurate?

[Anonymous Coward]

From TFA, Figure 4E:
The 99.3% error rate is a median of 449 single molecule read passes on 100 consensus sequences, which is approximately a 15-fold coverage.

Obviously, this single molecule method is already parallelized and multiplexed, and the output accuracy is 99.3%. Of course refinements and further read redundancy might lower the error rate.

I mean seriously, a group of brilliant people invent a new genome sequencing technology based on single molecule fluorescence and nanowaveguides, but you don't think they considered parallelization?

This method is incompatible with proofreading (to lower error read rate), because the read out is based on polymerase throughput activity. A polymerase is immobilized, lit up with two lasers, and when tagged nucleotides are processed, the waveguide catches a signal specific to teh wavelength of fluorophore labeled on each (of four T,A,G,C) nucleotides. There is no provision for immobilizing, lighting up, or monitoring any other proofreading proteins--doing so would require a major revision/re-invention of the technology.

Re:99.3% accurate?

[meatpan]

The reported accuracy rate of 99.3% for a single chip does not translate to 99.3% accuracy for an entire genome. Complex genomes are particularly difficult to measure at the edges of chromosomes and in areas of large redundant information (genomic 'repetitive' regions). The claim of 99.3% is stating that only the relatively easy to measure regions produce the reported levels of accuracy.

Re:99.3% accurate?

[MaryBethP]

umm...at a previous company our equipment used to cost $500k-$2M The equipment here is usually expensive and usually a large institute like Broad or Baylor would inevitably purchase it and then be able to charge people to run their sequences. A $5k pricetag would be per sequence run, so it would not necessarily be done for patients. Most of the work one on these systems is for research purposes--trying to find correlations amongst populations with disease to discover the genetic component, etc. A primary issue with the data that comes from this is where and how to store the data. 454 sequencers produce a terabyte of data (at full compression) for each run. When last I heard, they were only storing data for 2-3 weeks, as it cost less to re-sequence the data then it did to store it. There is an extreme need for a viable compression and storage solution and with a system that can sequence the entire genome, this will come to a serious head.

Re:99.3% accurate?

[Anonymous Coward]

i guess thats what they told you at school...

non-coding RNA has a function. In fact 40% of the genes are regulated by non-coding RNA. 88% of the genes are transcribed in the opposite strand and introns modifications, oh boy they can be horrible and cause alternative splicing or even frameshifts or splicing errors.

Codons are not 100% redundant, some triplets are better translated than others. 'Redundant' mutations can cause an alteration of tanslation speed and protein concentration.

And in situ hybridization is waaaaaaaaaay to expensive. Only used to locate certain sequence in a cell for instance or on a chromosome, not for error detection.

my 2 cents ;)

Re:99.3% accurate?

[Anonymous Coward]

i'd say more around 30-40 times with that error rate .

If it has a 4000 long sequence and 99.3, thats rubbish. 28 errors, you do the poison distribution...
  I dont want any error caused by accident on a single strand... come on, try analyzing that... Half of the sequences cant be mapped to the genome

Re:99.3% accurate?

[shaitand]

'There are good opportunities in the field of bioinformatics for computer scientists who know enough biology!'

While I've got a bio-engineer on the line... I'm starting to play a bit at home but I've found a bit a of a dead end. Perhaps you can clue me in on what procedures I need to Google next. The big problem I see is that while PCR itself is conceptually straightforward enough, and determining what the primers should contain is as well. There are entire documents online on designing primers.

But once you've determined the code for your primer what techniques are there to construct the primer itself? Preferably techniques that could be used on the small scale using home lab equipment.

Re:99.3% accurate?

[operon]

That is the reason why we sequence the same genome region more than once.

Re:99.3% accurate?

[Vornzog]

Home-brew genetics? Sweet stuff, and very trendy right now.

The equipment to actually build your own primers is pretty expensive, and I don't know of anything that is suitable for a regular university lab, let alone home brew. You'll occasionally find the equipment in a core facility at the local university.

With that said - basic, unmodified primers suitable for PCR can be had for relatively minimal expense on small scales, which should be enough to get you started. Search for 'oligonucleotide synthesis' and find the lowest cost per base. Add a 'protocol' to that if you want to see if you can set something up at home.

If there are universities or other resources near you, you may be able to get in on one of their bulk deals - the more primers you order, the cheaper.

Re:99.3% accurate?

[shaitand]

Thank you, just what I needed.

I've found $20/custom primer which leaves experimentation on a budget within reach.

Er

[Anonymous Coward]

You don't want to jump from 150 to 3 billion bases. Read up on shotgun sequencing. The mere fact that a given chunk is 150 bases is immaterial, though of course if you could lengthen that by a factor of five or ten it would improve accuracy and reduce computation on assembling the whole thing.

99.999% accuracy

[BigGar']

Headline you'll never see: Have your genome sequenced while you wait. No more than 30,000 error's or your money back!!!

Re:99.999% accuracy

[H0p313ss]

error's

That character you're using... I don't think it means what you think it means...

Re:99.999% accuracy

[troll8901]

... slowly leading to an era where helicopters can transform into airplanes, blood sampling is completely painless, brain contents can be downloaded by scanning the eyes, and clones can be produced in an afternoon.

Enter Arnold Schwarzenegger's clone, Adam Gibson!

Re:99.999% accuracy

[Anonymous Coward]

yes, yes, I was in a hurry and didn't proofread me preview.

Re:99.999% accuracy

[gnick]

error's

That character you're using... I don't think it means what you think it means...

You must be new here. You see, here on teh interweb, many of us are terribly afraid of word's ending in "s" - Plural's, possessive's, contraction's with the word "is", and occasionally even name's. The apostrophe is a polite way of warning the general reading public that an "s" is approaching so that we can brace ourselve's accordingly.

Re:99.999% accuracy

[squiggleslash]

' = "Look out! There's an "S" coming!"

Sub-$1000 genome sequencing

[morgan_greywolf]

Sub-$1000 genome sequencing will put the creation of 'designer' kids into the realm of the affordable for much of the middle class. Scary stuff. Now we just need to combine that with cheap and reliable cloning techniques and my plans for world domination will be comlete!

Re:Sub-$1000 genome sequencing

[Ethanol-fueled]

"...my plans for world domination will be comlete!"

Hopefully you'll fix that nasty intercalary deletion bug first!

Re:Sub-$1000 genome sequencing

[sam0737]

Sorry, it comes to my attention that you are missing a 'p' in the word 'complete', without a 'p', the world is never completed.

P, as in...I think you will figure that out by checking the spam.

Re:Sub-$1000 genome sequencing

[morgan_greywolf]

I will, but I gotta P first!

Re:Sub-$1000 genome sequencing

[CorporateSuit]

Hopefully you'll fix that nasty intercalary deletion bug first!

As long as it's not a missing G, T, C, or A, he'll be OK.

Re:Sub-$1000 genome sequencing

[Anonymous Coward]

Do you mean it is not NP-Comlete?

Re:Sub-$1000 genome sequencing

[cpricejones]

From what I hear, duplications are also bad.

Re:Sub-$1000 genome sequencing

[cpricejones]

From what I hear, duplications are also bad..

Re:Sub-$1000 genome sequencing

[Anonymous Coward]

I have nasal congestion, you insensitive clod!

Real-Time DNA Sequencing from Single Polymerase Mo

[mapkinase]

Abstract:

We present single-molecule, real-time sequencing data obtained from a DNA polymerase performing uninterrupted template-directed synthesis using four distinguishable fluorescently labeled deoxyribonucleoside triphosphates (dNTPs). We detected the temporal order of their enzymatic incorporation into a growing DNA strand with zero-mode waveguide nanostructure arrays, which provide optical observation volume confinement and enable parallel, simultaneous detection of thousands of single-molecule sequencing reactions. Conjugation of fluorophores to the terminal phosphate moiety of the dNTPs allows continuous observation of DNA synthesis over thousands of bases without steric hindrance. The data report directly on polymerase dynamics, revealing distinct polymerization states and pause sites corresponding to DNA secondary structure. Sequence data were aligned with the known reference sequence to assay biophysical parameters of polymerization for each template position. Consensus sequences were generated from the single-molecule reads at 15-fold coverage, showing a median accuracy of 99.3%, with no systematic error beyond fluorophore-dependent error rates.

Kicks ass on Moore's Law...

[djupedal]

> Company founder Stephen Turner estimates that such a chip would be able to sequence an entire human genome in under half an hour to 99.999 per cent accuracy for under $1000.

I think this qualifies as a true 'technological singularity' :)

Re:Kicks ass on Moore's Law...

[4D6963]

What does genome sequencing have to with Moore's law or the technological singularity?

Re:Kicks ass on Moore's Law...

[treddy]

You can also pretty easily show that our ability to sequence DNA is growing much faster than Moore's law. Right now, we seem to be in a nice world where we can process all the DNA we sequence, but we are already getting to the limits of pretty high powered workstations. The next step will probably tax the high powered cluster computers. But, assuming this rate keeps up, we quickly will reach a state where sequencing will be cheap and easy, and computer power will become the rate limiting step.

error correction

[bugs2squash]

Is there not some form of error-correction in the sequence itself that could be exploited ?

Something like the error correction on an audio compact disk ?

Re:error correction

[Hurricane78]

Yes. It's called "natural selection". :P

Re:error correction

[Anonymous Coward]

Well, the "error correction" on genes is somewhat contrary to the whole concept of evolution via mutation. Sure, there can be variation via reproduction, but if there were no mutations then the gene pool would remain stagnant.

As it is, "error correction" occurs when non-viable offspring is produced.

Re:Error Correction

[Anonymous Coward]

Each chip is a single use item. That's where the $1000 comes from ($1000 per use). So you couldn't use more time to run it several times on the same chip. You could run it on 10 chips though at a cost of $10k for the test.

This will be fun to see how insurance companies are going to handle it. "One chip is good enough"... heh, better hope the errors aren't important.

Re:error correction

[Anonymous Coward]

Yes.
http://en.wikipedia.org/wiki/DNA_repair

Only problem: enzymes work on a far slower time-scale than this sequencing technology.

Re:error correction

[rrohbeck]

No. Otherwise we wouldn't have evolution.
Interestingly, for most organisms the mutation rate has adjusted itself such that it is fairly optimal for the environment. Viruses, e.g. HIV, have totally sloppy copy methods while more complex animals like mammals make far better copies - the longer lived the better.

Re:error correction

[Mab_Mass]

Is there not some form of error-correction in the sequence itself that could be exploited ?

Well, not really. The good news, though, is that it isn't a problem.

In nature, error correction works through the fact that during DNA replication, there is always a "mother" strand and a "daughter" strand. As the DNA sits in a cell, methyl groups are added to the backbone of the DNA, which guarantees that both of the mother strands are methylated. (Don't worry about the details here - just think of this is a boolean variable tagged on each strand.)

If an error occurs during replication, the DNA double helix will be disrupted and this abnormal secondary structure is detected by the repair mechanisms in the cell, which chew out the bases from the new strand (using the methyl group tags to identify which is which), and re-synthesize it.

When trying to sequence in the lab, each particular read occurs only once and there is no way to be completely sure that there are no errors. The exact nature of these errors varies wildly across different sequencing technologies, and I'm not even going to try to be exhaustive in describing all of them. In the end, though, it doesn't matter, since when anyone is trying to sequence an organism, they don't sample each region only once. Instead, sequence is only considered to be accurate once it has 5-10X coverage of reads for the same region. (Reading from different molecules from the same part of the genome from different cells.)

Because of this redundancy, the final sequence tends to be very accurate, regardless of the technology used to generate it. The noisier the technology, the higher the coverage needed for the same level of accuracy. In the end, it all comes down to cost - how much are people willing to pay for what accuracy of how much coverage.

The really great thing about what Pacific Biosciences is doing with this new technology is that the time and costs are dropping to be so low that it will no longer be an issue. That is, assuming that they can make it work as well as the papers claim in a reliable manner. If so, a lot of biotech companies (including the one I work for) will be out of business. Much of biotech is centered around finding quick, cheap ways to get at the information, but if you can suddenly get a whole genome sequence for the same time and cost, many of these older technologies will become obsolete, almost overnight.

The flip side for someone with computer skills, though, is that will this sudden surge of data, there is going to be a huge need for people with the ability to make sense of it all.

Re:error correction

[Anonymous Coward]

There's two strands. You sequence both. If they don't match, you screwed up. Do that part again.

slashdotted!

[Anonymous Coward]

can't slashdot automatically mirror links created in articles? that way they are always readable... slashdotted links are annoying...

Re:slashdotted!

[Anonymous Coward]

There are probably copyright laws against this. Google news has legal problems just when copying the summary

current read lengths... a small erratum

[Anonymous Coward]

Several "next generation" sequencing methods currently produce short sequences, or "tags". 454 sequencing isn't one of them - 454's typical read length is a few hundreds.

MOD UP

[Anonymous Coward]

see title

Comment removed

[account_deleted]

Comment removed based on user account deletion

Bad summary

[PeterPlan]

Using 454 sequencing you get average read lenghts of ~400-500 bp. Read lenghts around 20 bp would be pretty much useless. At least for de novo sequencing..

Re:Bad summary

[damn_registrars]

Using 454 sequencing you get average read lenghts of ~400-500 bp

I suspect someone had confused 454 with the other popular next-gen sequencing technique from Illumina, which does give very short reads.

Read lenghts around 20 bp would be pretty much useless. At least for de novo sequencing..

Not necessarily. If you can drive the cost/base down far enough, you can make short reads worthwhile if you use a shotgun approach and try for large-scale coverage. Especially if you can produce the short reads at a lower rate of time/base.

Re:Bad summary

[rnaiguy]

The Solexa sequencing platform has been VERY useful (admittedly not so much for de novo sequencing), with a read length of ~30 or less.

Re:Bad summary

[nodrogluap]

The use of short reads for de novo assembly only makes sense if you want a rough draft of a genome, not the complete thing. There are way too many transposable elements, repeats, variation, etc. to accurately reconstruct even a bacterial genome with short reads. Nowadays, people don't even bother trying to piece it all together. They get down to a few dozen large fragments and say "good enough". It just costs too much to get the last 1-2% with a random sequencing approach.

Re:Bad summary

[damn_registrars]

The use of short reads for de novo assembly only makes sense if you want a rough draft of a genome, not the complete thing. There are way too many transposable elements, repeats, variation, etc. to accurately reconstruct even a bacterial genome with short reads.

At the conference "Intelligent Systems for Molecular Biology" (ISMB) this year there was a discussion on exactly that. I can't say that there was an absolute consensus on how to do it, but there was fair acceptance of actually doing a hybrid approach:

• Random short reads (illumina method) for cheap coverage of most of a genome
• Directed sequencing via the 454 method for difficult regions (like the ones you described)

Basically, the 454 method, while costing significantly less per base pair than the older sanger method (and much, much faster than the yet-even-cheaper method of having [grad] students read bases off agarose gels), is still more expensive in terms of cost-per-base than illumina.

And of course in some cases the interest may be entirely in the "coding" region of a genome, such that the rest may not be worth its additional cost anyways.

Link: Google Cache version

[Anonymous Coward]

(from Google Cache) Reading DNA sequences from single molecules of polymerase using nanotechnology

By Neruos

[Anonymous Coward]

Fact: .01% is enough to cause mutation.

Article in Science

[prograde]

I assume that the hardware at Science can withstand a slashdotting better than the crappy blog linked in the summary:

http://www.sciencemag.org/cgi/content/abstract/323/5910/133

GATTACA is here

[arbies]

Does anyone remember the movie, "GATTACA"?

Re:GATTACA is here

[oni]

Gattaca was supposed to show us a dark future. It was supposed to be a cautionary tale. The message was, "if your DNA isn't good enough, you'll have to make do banging Uma Thermon - poor you."

I don't think the producers thought their cunning plan all the way through.

GATTACA here we come!

[greg_barton]

1/2 hour for $1000, eh? And in another 5-10 years we'll cut that in half or more, both time and cost. It looks like the instant gene sequencing tech from GATTACA will be with us in most of our lifetimes. But even with this announced breakthough it'll be functionally the same.

Battle Tactics :)

[headkase]

Reminds me of that "Lost in Space" or whatever it was called remake. Terrible movie but remember the scene where they are fighting the spider-things and they slap down a chunk of one onto a machine which pretty much instantly reconstructs the full organism and then goes on to suggest ways to fight it based on how its built? Yeah, this could lead to one of those machines being reality.

Re:Battle Tactics :)

[timeOday]

This is entirely reasonable and desirable if you replace "spider-thing" with "cancer" or "aids," or even "common cold." Gene sequencing your disease and taking the right medicine for what you *actually* have - instead of today's educated guesswork - will be a HUGE advance. Thousands die every year because they have to guess a year in advance which flu strains will be prevalent and usually guess wrong.

I guess I can drop my X prize plans

[damn_registrars]

Since this technique should be a shoe-in for the Archon X Prize.

Re:+digg

[Anonymous Coward]

[citation needed]

Old News

[Anonymous Coward]

I work in this field (our lab has 2 Illumina sequencers) and unfortunately the article is slashdotted. The single molecule stuff the summary seems to be talking about is from Pacific Biosciences (with a few hundred/thousand pools). These guys are well funded - their backers include Kliner Perkins Caufield and Byers [sp?] (a historic VC firm) and I'm hearing stories how they're turning down offers for more funding. They're probably going to be the first of the 3rd generation sequencing technologies. Next-gen sequencing has been in the news a lot lately - if you have access, check out some of the recent sequencing papers in Nature and Science. In any case, there are quite a few competing technologies. There seems to be a lot of talk about different error rates, but in reality the error introduced in the sequencing really depends on the technology. 454 aims for longer reads at the cost of fewer reads - the data I've seen gives a few hundred thousand 250-350 base reads per run of the machine. The Illumina sequencer we use gives us short 36(now longer) base reads, but we get 60 million of them. There are also other technologies which use similar fluorescence like the SOLiD system and the open-source Polonator developed in George Church's lab. All these 454, Illumina, AB SOLiD, etc have been out for at least 18 months now.

Re:Old news

[Anonymous Coward]

2nd January 2009
http://www.sciencemag.org/cgi/content/full/323/5910/133

ForeSight.Org Down

[phantomcircuit]

I guess they didn't have the foresight to use a real host.

I'll be here all night folks try the steak.

not for cloning, but therapy

[planckscale]

I believe the true benefit of this technology will not be for cloning, but for general medicine. For example, you would go to the doctor with a lump, and instead of him doing a biopsy, find cancer, chemo, invasive surgery etc etc, they first take your DNA, sequence it and then take the biopsy and identify the origin of the cancer (is the lump actually metastasized from your pancreas?). Then work on resolving the cure just based on your genetic makeup, rather than a shotgun approach. Additionally medicines for genetic problems, and a number of other diseases would be custom-tailored for your genetic makeup. If you are prone to hypertension, your DNA sequence could prove if you carry the genes for that malady. Really what this is about is a revolution in medicine. It's a private company now that is snatching up all the biggest heads in silicon valley - if and when this goes public, it could be an amazing investment.

Slashdotted: Abstract and Fulltext

[chihowa]

It looks to be inaccessible. Here are the abstract and fulltext links.

Re:Slashdotted: Abstract and Fulltext

[Anonymous Coward]

It looks like wikipedia has a brief blurb about this technique as well

DOI was Re:99.3% accurate?

[Anonymous Coward]

DOI: 10.1126/science.1162986

finding out what we're really made of

[Anonymous Coward]

unfortunately for some, that will prove that our spirit is our outstanding feature.

maybe 60 to 1000 are significant?

[peter303]

Forensic genetic identification currently uses about 60 important genetic markers. Thats good enough to convict in a court law since the the chance of a duplicate may be less than a billion to one depending on marker combination.

Although humans differ from one another in about 0.1% base pairs for a total of 3 million, the number of difference that describe human variability may be vastly smaller than this. First you discard non-coding DNA which gets you done to 30,000.

Re:maybe 60 to 1000 are significant?

[nwf]

Although humans differ from one another in about 0.1% base pairs for a total of 3 million, the number of difference that describe human variability may be vastly smaller than this. First you discard non-coding DNA which gets you done to 30,000.

Except that when our differences are so small, the non-coding regions are even more important. They control what genes are active and to what degree. That's nearly as important as the genes themselves.

Genes are only part of the puzzle. You need to know what to do with them, and non-coding regions provide some of that along with the cellular machinery.

Scientists used to call them "junk" DNA where junk == "I can't figure it out". Why would cells spend all that energy maintaining something useless? Not very likely.

Re:maybe 60 to 1000 are significant?

[Anonymous Coward]

from what i've learned from programs like nanopond, some of our dna exist just to make harmful mutations less likely.
And some is just some old stuff that give no disadvantage (or advantage ) by keeping it.

Old news

[Anonymous Coward]

The only Science article on this topic was published in 2006: http://www.sciencemag.org/cgi/content/summary/311/5767/1544. It won't be new news again until the product ships - supposedly in 2010

piggy backs on Moore's Law

[peter303]

Shotgun sequencing depends heavily on supercomputer. Thats a thousand-fold every 15 years right there. Multiply that by more intelligent software, understanding of genetics, and sequencing hardware, you may be squaring that rate.

error rate of 99.9 is standard min. for research

[Anonymous Coward]

Typically the minimum goal is 1 error / 1000 bases for sequencing a new genome. Current sequencing has a much higher error rate though, so each region must be sequenced ~5-7 times to reach that goal. The human genome is at about 1 error per million, last I heard (goal is to decrease that to 1 error / billion by resequencing repeatedly). There aren't many good reasons at this point to sequence entire genomes for individuals, since there are better ways to test for genetic diseases. But for researchers this is very promising.

re the error rate thing and over view

[cinnamon colbert]

This is just an historical accident; 99.9% was what could be done with what people judged "reasonable" effort and cost a few years ago; unless you know whatyou are going to use the sequence for, you don't know what error rate is acceptable

There are medical test that rely on dna sequence, eg myriad makes a fortune from sequencing the gene that gives women hereditary breast cancer. I don't know what there claimed error rate would be, but that would show you what is acceptable in todays clinical market place.

As for the "de novo" rate, eg the difference between a child and its parents, or between two identical twins - I don't think this has been accurately measured, but I do believe that single base changes (eg AATTC to AAATTC) are not as comon as insertion or deletion of several bases..which goes to show that biology is ocmplicated beyond belief

on the overview side, for those of us who follow this, pacbio has been hyped beyond belief, and the production of actual data eagerly awaited; time will tell if single molecule sequencing is to be the wave of the future, helicos bioscience (HLCS) has had a very hard time selling its system, which has been on market for about a year now; for instance, one problem is "blinking" single molecules that are fluorescent can go into a "dark" state where they don't give a signal - which is kind of an obvious show stopper

Re:re the error rate thing and over view

[Mab_Mass]

Keep in mind that part of why Helicos has had such a hard time is that their read lengths suck, especially compared to the 454, which generates comparable throughput for most applications. They were a classic case of a technology that got to the market too late to have a great impact since other technologies had already surpassed them.

nitpick

[Zenaku]

One base-pair does not a gene make.

Re:nitpick

[TheMeuge]

One base-pair does not a gene make.

But a one base-pair change can unmake the gene pretty well.

Tons of major debilitating mutations are due to a point mutation.

Re:nitpick

[jamstar7]

One base-pair does not a gene make.

But a one base-pair change can unmake the gene pretty well.

Better question: How many base-pairs does it take to unmake my hot 19 year old neighbor's jeans?

first, this is not news!

[Koppology]

While they may have only recently published the article, people in bioinformatics have been going crazy about Pacific Biosciences for at least a year.

I recently went to a series of talks on Next Generation Sequencing, and there was an interesting chart that showed that when you factor in sequencing cost, read length, and accuracy, high throughput sequencing is actually *outperforming* Moore's law by a factor of 5 or so!

Regarding the error rate, just a few years ago, 454 had error rates of almost 5% but with redundancy it became negligible. Since then, error rates have gone down dramatically.

Also, an Anonymous Coward up there is wrong -- (0.993)^3 is *less* than 0.993. It should be (1 - 0.993) vs. (1 - 0.993)^3. I don't know why it's been modded to informative. Check your math!!

Grammar ambiguity

[nsayer]

Company founder Stephen Turner estimates that such a chip would be able to sequence an entire human genome in under half an hour to 99.999 per cent accuracy for under $1000

Does that mean that the chip costs $1000 or that each human genome processed costs $1000?

Re:Grammar ambiguity

[Hatta]

Both. Each chip processes one human genome.

Re:Grammar ambiguity

[Anonymous Coward]

GP meant: "will the chip itself cost $1000, or will it cost $1000 to sequence an individual's DNA using the chip?"

(Hint: There's more than an order of magnitude difference in cost to the patient at the hospital. If the chip only costs $1000, then it may only cost $50-100 to get your DNA sequenced.)

Re:Grammar ambiguity

[nsayer]

I think GP's reply implied that the chip was single-use - that is, a chip has a useful lifespan of a single act of sequencing. In which case, he's right that both the chip and a single sequencing costs $1000.

And I thank him for answering my question. :)

4000 bases, with a 99.3% accuracy

[Anonymous Coward]

99.3% of your bases are belonging to us!

Re:4000 bases, with a 99.3% accuracy

[4D6963]

99.3% of your base are belong to us!

Fixed it for you..

Ask!?! Re:99.3% accurate?

[Anonymous Coward]

[Suspects] that come up positive can ask for the more accurate test.

Umm... kind of like getting a lawyer for free if you need legal representation and lack funds, if you come up positive, it should be the default that they run the more accurate test.

It's getting deep in here.

[Anonymous Coward]

It's common practice on Slashdot to read the article before posting.

You're new here, aren't you?

Actual link to scientific article

[onco_p53]

http://www.sciencemag.org/cgi/content/abstract/323/5910/133

Abstract:

We present single-molecule, real-time sequencing data obtained from a DNA polymerase performing uninterrupted template-directed synthesis using four distinguishable fluorescently labeled deoxyribonucleoside triphosphates (dNTPs). We detected the temporal order of their enzymatic incorporation into a growing DNA strand with zero-mode waveguide nanostructure arrays, which provide optical observation volume confinement and enable parallel, simultaneous detection of thousands of single-molecule sequencing reactions. Conjugation of fluorophores to the terminal phosphate moiety of the dNTPs allows continuous observation of DNA synthesis over thousands of bases without steric hindrance. The data report directly on polymerase dynamics, revealing distinct polymerization states and pause sites corresponding to DNA secondary structure. Sequence data were aligned with the known reference sequence to assay biophysical parameters of polymerization for each template position. Consensus sequences were generated from the single-molecule reads at 15-fold coverage, showing a median accuracy of 99.3%, with no systematic error beyond fluorophore-dependent error rates.

New Scientist Write Up

[Fnord666]

Here is an article in New Scientist about the new process. It explains it fairly well and even defines what a ZMW is.

Error Correction

[LUH 3418]

Many people seem concerned about the reading error rate. However, as it's been pointed out, it should be easy enough to read a DNA sequence multiple times (or read the whole genome multiple times) to decrease the error rate significantly. If you have one chip that can read the entire human genome in 30 mins, you can have the same chip read it twice in an hour, or four chips reading four copies in 30 mins.

Furthermore, if you're using a technique like this to map a person's genome, you can be clever about it. Base pairs code genes, which is something you can take into account. For example, if you're reading the eye color gene, and your machine somehow consistently makes mistakes in that area, you can compare your reads to the few possible known eye color genes, and pick the most likely based on the genetic sequences of the entire gene.

Getting closer to...

[Anonymous Coward]

G A T T A C A?

454 Sequencer isn't ~that~ bad.

[cerebis]

The 454 pyro-sequencer currently produces 400bp reads, not 20bp. Granted, that's still a fair bit shorter than this experimental tech claims, but it's also a commercialized product you can actually buy right now. I think it would only be fair to quote current performance figures.

Re:454 Sequencer isn't ~that~ bad.

[lbbros]

We have one of these beasts where I work. It's tremendously expensive to run (~$5K for a single run, although you sequence 40 million bases in 200-400 bp reads), and the most daunting part is the data analysis. So far there are three people that just work on what that sequencer crunches, and it took quite a while, time-wise, to develop efficient workflows.

Re:+digg

[Anonymous Coward]

Larry bagina, is that you?

Single Molecule Sequencing Rocks

[treddy]

One of the real advances here is the ability to do this on a single molecule. Existing DNA sequencing techniques all depend on an amplification step, known as ploymerase chain reaction (PCR), in which the DNA is iteratively duplicated (this is done by basically hijacking DNA replication machinery from bacteria). However, PCR introduces numerous biases in the final population of DNA molecules: shorter segments and certain sequences are easier to duplicate than others. As a result, what you end up sequencing is always skewed. This may not be too important when it comes to (re)sequencing a genome, but there are a whole cadre of experimental techniques that use sequencing to investigate regulation and modification of DNA, and here that bias can really skew findings and generate many false positives (things that amplify too easily) and negatives (things that don't amplify well at all).

Where's the problem, exactly?

[Anonymous Coward]

With a CD, the error you are correcting for is an error on the CD, not in the CD player. With sequencing, it's not the DNA code that fails - it's the machine you're using to read the code that fails. The solution is similar to Reed-Solomon in that you sample every DNA region multiple times, so that you can ignore the "lost data points" of misreads. There is no bulit-in mechanism like that for DNA as a strand alone. General DNA proofreading happens during replication and is done by polymerases that are doing the copying in the first place; there are a slew of other specialized mechanisms for DNA damage repair but they're again separate enzymes and not inherent characteristics of the DNA.

Re:cost of sequencing is a reasonable determinant

[CrankinOut]

If the article had stated that the cost were $1,000,000 to do the sequence, then the potential applications of the technology would be severely limited. Getting the cost (not the charge for the service) down creates the opportunity for more studies to be performed, more financial accessibility for patients, and less resistance for insurance companies or Medicare to deny charging for the study when it's indicated.

In medicine, the cost of a study, as well as its reliability, availability, and predictive value, enters into the decisions made in clinical management.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Reading a sequence is not the same as creating one

[CrankinOut]

Just as seeing the moon doesn't require the same amount of effort as landing on it, reading a DNA sequence doesn't mean that selective modification is "just around the corner."

Real applications of this, however, include looking for gene sequences in adults which predispose them to diseases (e.g. breast cancer) and then providing counseling and monitoring commensurate with that risk, a far less expensive effort than monitoring everyone for the same disease, even if they aren't at risk. Also, one could use this on embryonic cells obtained through amniocentesis to screen for hereditary diseases is families where there are risk factors.

Sig Figs and Multiplication

[mosb1000]

That's not how significant figures work with multiplication. .007 has only one significant figure (the 7) .007 * .007 = .00005 which also has one sig fig (5). 1-.00005 = .99995, since the 1 is a known integer value (and therefore has an infinite number of significant figures).

That being said, the parent has an incomplete grasp of statistics in this case, and using this formula is inaccurate. However, I am at work, so I will not do an in-depth statistical analysis here.

Finally!

[Locke2005]

Now I can make sure that hot chick isn't my long-lost sister BEFORE I sleep with her! And save myself the Luke Skywalker type embarrassment of almost having done my sister...

However, don't you just HATE it when your date asks you, "Before this relationship progresses any further, I'll need a sample of your DNA!"

$1000?

[Anonymous Coward]

I'm sorry, but I only pay $2000 dollars an hour for one thing. And I don't even really pay that much for that.

Re:Real-Time DNA Sequencing from Single Polymerase

[Anonymous Coward]

I do not believe in karma.

Yet here you are, karma whoring to pay for your kids' schooling.

It does not rock

[pnotequalsnp]

Considering current sequencing technology generates terabytes of data per day (see the Sanger center), then wouldn't it be efficient to maximize the amount of information per pixel (i.e. per byte)? This method is actually is much worse (orders of magnitude) than the current method. There are many other problems with what they do, but hopefully the cash infusion can last them another 2 years until the write a paper like this. BTW, the say that appropriate camera tech. will be available in 2-5 years, but they're ready now! They might be buying time...

Reading words at a time ...

[Anonymous Coward]

... spells dyslexia.

Re:Reading a sequence is not the same as creating

[morgan_greywolf]

Just as seeing the moon doesn't require the same amount of effort as landing on it, reading a DNA sequence doesn't mean that selective modification is "just around the corner."

Who said anything about 'selective modification'? Read what you wrote:

Also, one could use this on embryonic cells obtained through amniocentesis to screen for hereditary diseases is families where there are risk factors.

If you can screen for hereditary diseases, you can also screen for desired traits such as hair color, eye color, propensity for obesity (yes, there's a gene for that), intelligence, etc.

In no time flat, we'll have become a race of athletic, attractive, social and congenial people who all get along, but are all little short in the intelligence department.

compasses

[Anonymous Coward]

Compasses are usually not a big problem at sea--if you wait a bit you can usually see the stars or sun and know N/S.

The problem at sea is time. It is vital to know exactly what time it is so you can know your longitude.

Darwin's vessel, the "Beagle" was a mapping exploration vessel and carried 22 chronometers.

I'm not sure how that works, since I've noticed that these days, when we have the ability to generally 'synchronized time signals, and everyone has at least one and usually 2 or more clacks available, the probability that any 2 of them agree seems pretty low.

Of course, for most everyday life we could do fairly well by knowing time to the 1/4 hour. Excess precision available does not mean that it is desirable.

wizodd

Note To The Slashdot Editor

[TheVoiceOfEnigma]

Before slashdot posts a science story it should check to see if the submitted story is accurate. The Ti pyrosequencer by 454 has a Q20 read length of 400 bases, not 20 bases. The 99.3% read accuracy described in the story is at 15-fold coverage. This means that the raw base accuracy (sequencing a template once) is much lower than 99.3%. This is in comparison to the Applied Biosystem's SOLiD instrument, which has a raw base accuracy of excess of %99.94.

enjoy

[Anonymous Coward]

454 has read length of 250 and the GS FLX titanium, which is the upgrqde, has a 400 length read.

Accuracy is not an issue, just do multiple coverage, even on 454 30-40x coverage is the standard.

20 long reads is on the Illumina. Shorter but more reads and cheaper/base.

Article is rubbish just as the comments below the article. Has anyone of you smart biologists ever seen a DNA sequence?

Upon request, i can send you the dataset from 1 454 run. (kidding, of course not, intellectual property, and it is too big for your littel brrains to analyse).

And secondly the pacific biosciences article has been published more than a month ago, so you guys are late. 20 nov to be correct: http://www.ncbi.nlm.nih.gov/sites/entrez?db=pubmed&cmd=Retrieve&list_uids=19023044

btw the golden standard is the Helicos now. But toooo expensive 3 mil whereas a 454 or Solexa is around 500k; Enjoy

exagerated improvement?

[Anonymous Coward]

but who wrote this and how accurate is it - 454 has been giving reliable 220 to 240 nt reads for over a year (done 2 genomes this way at 50x coverage) and the new version does 400+, the new cells do many more simultaneous reads. Even solexa new modules do 40+ nt. So, while this sounds like serious progress (not sure yet - need to read the article) - its not quite the orders of magnitude suggested by the person posting this.