IPF 2010: DNA sequencing: Faster, faster, ever Moore

Listening to Suzanne Wakelin of Illumina Inc, who gave a talk on Monday, 25 October, titled ‘From Photonics to Genomics: Lasers and Imaging Technology Enables Next-Generation DNA Sequencing,’ I was awestruck by something she said about Moore’s law.

That law, you will recall, is named after Intel cofounder Gordon Moore, and to my mind, it has nothing at all to do with DNA sequencing. The law, which Moore formulated in the 1960s, basically states that the number of transistors on an integrated circuit doubles every two years, a rate that has profound implications for the size, speed, and price of processors.

But despite that somewhat narrow definition, Moore’s Law has become something of a yardstick for measuring growth and progress in other fields. People sometimes discuss the increase of energy-storage capacity in alkaline and lithium batteries in terms of Moore’s law—usually to underline the fact that battery growth falls far short.

What Wakelin (shown above) said about DNA sequencing trended in the opposite direction. The exponential growth of sequencing has far outpaced Moore’s law in the last of couple years, and it shows no signs of abating. She showed a semi-log plot of transistor milestones in the history of computer processing, very much like this one from Wikipedia . But Wakelin’s plot also showed (on a different y-axis) the rapid rise of sequencing, which is steeper than the standard Moore plot.

The first complete genomes of complex organisms were being sequenced and published in 1995, just when I was starting graduate school. Although it sounds strange to say, in terms of DNA sequencing, those first milestones are ancient history—thanks to the accelerated, greater-than-Moore growth of the field.

One way to look at the growth is in terms of cost. Among the figures that Wakelin reported were the associated costs of some of the notable past human genomes. The original human genome project was completed in 2003 after more than 10 years and a final price tag of about $3 billion.

In 2007 James Watson’s DNA was sequenced at a cost, according to Wakelin, of a mere $2 million. (New Scientist gives the price tag for cracking Watson’s genetic secrets as $1 million, but the two estimates may not actually be at odds, because of different assumptions about computing power, technician time, and other costs.

In 2009 the cost of sequencing genomes was in the $100 000–$200 000 range. In a September press release the National Human Genome Research Institute (NHGRI) put the cost at less than $40 000—something the institute is determined to change, according to its statement, ‘Ultimately, NHGRI’s vision is to cut the cost of whole-genome sequencing of an individual’s genome to $1000 or less, which will enable sequencing to be a part of routine medical care.’

In her talk Wakelin predicted that the field would hit what she calls the ‘holy grail’ of $1000 genomes in the next couple years. Despite similar past claims , the rapid growth of sequencing would seem to support her optimism.

Base pairs

Taking together some of the data points, the cost of DNA sequencing now seems to be dropping by an order of magnitude a year. But what about the number of pairs of DNA sequenced? How does that measure up to Moore’s law?

According to the Department of Energy , 200 million base pairs were sequenced for all genome projects in the whole of 1998. By 2003 one large project alone, the DOE’s Joint Genome Institute (JGI), sequenced some 1.5 billion bases in a month.

If the march of DNA sequencing had been increasing according to Moore’s law, then, based on my back-of-the-envelope calculation, the JGI’s capacity would not have reached that 1.5 billion bases milestone until next year at the earliest. In fact, the JGI lists its likely 2010 sequenced output at 1.1 trillion bases—more than four times the 256 billion bases it sequenced in 2009, which was itself more than five times the 42.8 billion bases it sequenced in 2008. Genome data are being accumulated at a rate far, far faster than Moore’s law.

That makes for astounding comparisons between the recent past and today’s state of the art. According to 454 Life Sciences , one of the major players in the DNA sequencing field and now part of the pharmaceutical giant Roche, the company’s sequencers can decode 100 million bases in a single business day—fast enough, it says, to solve the human genome in about 10 days if the sequencer runs for 24 hours a day during that time.

Considering that the first human genome took more than a decade to solve, that is an incredible advance indeed. But where does it stand in terms of Moore’s Law?

Again, based on another of my back of the envelope calculations and on the DOE sequencing numbers, comparing the amount of DNA sequenced in 2010 with what was done just two years ago would be like comparing the computer you have today with the one you were using in 2000.

According to a feature article in this week’s Nature, the acceleration is set to continue at an even more breakneck pace. The article, timed to accompany the first major publication of the 1000 Genomes Project , points out that while there were only two human genomes solved a decade ago and only a handful as recently as last year, estimates of the number of human genome sequences currently in the works are jarringly large: ‘Although far from comprehensive, the tally indicates that at least 2,700 human genomes will have been completed by the end of this month, and that the total will rise to more than 30,000 by the end of 2011.

So, measured in terms of the number of human genomes (I’m not sure how that scales exactly in terms of the number of DNA bases, since there are lots of other organisms sequenced every year), the current pace of sequencing is increasing 10-fold or more per year.

Another of my back-of-the-envelope calculations suggests that comparing the number of genomes solved today with those solved in 2008 would be more like comparing your computer today with a 386 processor running whatever came before Windows 95—sadly, the exact sort I was using when I started graduate school.

Jason Bardi

All the talks at IPF 2010 were recorded and are now available on video .