1 machine, 4 weeks now enough to sequence human genome
arstechnica.com
arstechnica.com
"Fragments this small probably can't be assembled into a complete genome, but the authors were able to compare them to the reference human genome maintained by the National Institutes of Health. They were able to align sequences totaling 2.5 billion bases to the published version of the human genome."
So, it's apparently now possible to generate redundant, small-fragment coverage of the human genome in four weeks, but that's a long way from a full, assembled genome. The researchers had to cheat by performing alignments of the tiny fragments with the known, assembled human genome sequence. This was the same game that Celera Genomics had to play to claim a "fully assembled" genome from shotgun sequencing -- but they were using larger reads, and still failed to be scientifically convincing.
This is interesting technology, but the title is a massive exaggeration.
...the HeliScope generates DNA sequence data as a massive collection of very short reads - but while the Illumina platform is now routinely generating reads over 100 bases long, the HeliScope generates reads on average just 32 bases long, with only a tiny fraction exceeding 50 bases in length. In fact, the reads are deliberately filtered to exclude any extending for over 70 bases, as these are highly enriched for technical artefacts.
Stitching together a genome sequence with such short reads is a substantial challenge, especially in regions where the sequence is repetitive - and indeed the technology can only cover 90% of the reference genome compared to 99.9% for a genome recently sequenced to similar depth with Illumina.
http://scienceblogs.com/geneticfuture/2009/08/first_human_ge...
So they can still stitch together 90% of your genome within a month, and there's no reason to think this will not improve in the future.
As I understand it, there's enough repetition and duplication in the genome that many of these little lengths are going to be identical: AAAA...AAAA or GGGAGGGA...GGGA etc. and that these little strips will match in many different places when placed alongside a fully assembled reference genome. They will also match and overlap a lot with each other. This means that you can't tell where they're supposed to go when you try to re-assemble your collection of little bits into a complete sequence. This means that you're always going to be left with a collection of bits left over that you have no way of placing into the sequence, hence less than 100% coverage.
I guess you would need to be using reads long enough to ensure that each of your fragments was unique to get 100% coverage.
It's a very important step for the PR department of Helicos.
Unless they can dramatically increase the read length or reduce the error rate, this technology isn't going to be very useful. The genome is not randomly distributed, and there are lots of repeated elements that make assembly impossible with such short fragments of sequence. Even shotgun sequencing -- which has read lengths of hundreds of bases -- isn't good enough for the human genome.
That said, the technology might be useful for something like large-scale SNP analysis, where (even though they've got error rates >1%), they could sequence the same genome dozens of times, and get an idea of the interesting variations. Maybe. But the current technology isn't very useful for genome sequencing and assembly.
The main problem with DNA sequencing is that nobody has really commercialized a way to "passively" read the sequence. All of the currently marketed technologies involve some kind of chemical fluorescence process where little bits of DNA are used as templates. The templates are extended base by base with fluorescent-labeled bases. Between the chemical reaction, and the need to accurately measure the fluorescence, you can really only go so fast and maintain accuracy.
Once you do the chemistry and have the raw data, you've really just got TONS of tiny sequences (known as "short reads"). These reads are anywhere from 20 - 30 bases or so, depending on the technology. They need to be assembled into a complete genome by figuring out where they overlap with each other. This is a computationally expensive problem, but it's reasonably doable with a small Linux cluster.
This is a huge improvement in sequencing (most notably, it removes a lot of human intervention). That said, large-scale sequencing is still in its infancy. In my opinion, all of this is a stopgap until someone is able to engineer a process that gives really long reads. There are numerous problems with assembling short bits of sequence. Most notably, there are lots of areas in the genome which have a huge amount of repeated DNA sequence. Since reads from neighboring parts of the repeated sequence can be identical, there's often no way to know how many are duplicated. Pacific Biosciences claims they can sequence 2.5 kilobases of a single molecule with their technology (not yet on the market). Something along those lines could really change the equation.