How Low Can We Go? Molecules, Photons, and Bits

The dramatic decline in the cost of sequencing whole human genomes over the past five years surprised even the most optimistic scientists (except possibly George Church). Starting from Roche 454’s $1 million Watson genome in 2006, we are now sequencing genomes for a few thousands of dollars each.

The question is: how low can we go? Can the current generation of fluorescent chemistry sequencing systems be improved to produce medical-quality whole genomes for $1,000? $500? $200? Or do we need to replace current technologies with new approaches, such as nanopores or electron microscopy?

I’m a physicist by training, so to answer these questions I look to the fundamental physical units of sequencing genomes: molecules, photons, and bits. The three major costs of DNA sequencing are the reagents (molecules), sequencing instrument (photons), and computing (bits). In our sequencing system, each of these cost elements is already less than $1,000 for a whole human genome (6 billion bases, 40x depth, >95% coverage). Let’s see how low we can go.

Molecules. A powerful way to reduce the costs of molecules (mostly enzymes and fluorescence-tagged nucleotides) is to reduce the volume of reagents needed to sequence each base. We use miniaturization techniques borrowed from the semiconductor industry and currently achieve a volume per base of 700nm x 700nm x 50um. We are working to reduce this volume to 250nm x 250nm x 10um. Achieving this volume reduction could reduce the cost per base by about a factor of 40, from < $1,000 to ($1,000 x (250/700) x (250/700) x (10/50)) or approximately $26 per genome.

Photons. The cost of photons is the cost of the optical and fluidic instrument designed to generate and capture photons from the fluorescent molecules. We can reduce the instrument cost per genome by successfully using more, faster cameras. Our current instruments are equipped with two electron multiplying charge coupled device (EMCCD) cameras. There is a new generation of fast complementary metal oxide semiconductor (CMOS) cameras, developed for other industries that are about 15 times faster than our current cameras (and also less expensive). New sequencing instruments that successfully use four of these fast new cameras could reduce the instrument cost per genome by about a factor of 30, from < $1,000 to $1,000/(2 x 15) or approximately $33 per genome.

Bits. The cost of bits is determined by the amount of processing and storage needed to reconstruct a whole human genome from billions of measurements. Moore’s Law (i.e., computing hardware costs go down by a factor of two every 18 months) delivers computing hardware cost reductions. Our exclusive focus on sequencing human genomes permits us to standardize and optimize our software, and should also enable continuous software cost reductions. Achieving these hardware and software cost reductions could enable a reduction in total computing costs by about a factor of three every 18 months, which means < $1,000 per genome today could become approximately $25 in five years.

These are the major costs. The other costs (sample preparation, labor and overhead, etc.) may be reduced by economies of scale due to the standard, repetitive nature of whole human genome sequencing. An automated factory that can sequence thousands of genomes in large batches might be able to reduce these costs to tens of dollars per genome.

So to answer the question of when do we have to abandon the existing fluorescent chemistry sequencing systems and move to a new generation of sequencing technology is: I don’t believe we have to. Improvements to existing systems provide the potential to reduce whole human genome sequencing costs to a few hundreds of dollars, then a few tens of dollars, in a relatively short period of time. None of these big cost reductions require major technology breakthroughs. They result from good engineering practices, economies of scale, and powerful tools borrowed from other industries, notably semiconductors.

A recent article in JAMA (Pasche B, Absher D Whole-Genome Sequencing A Step Closer to Personalized Medicine (Editorial). JAMA. 2011;305(15):1596-1597. doi: 10.1001/jama.2011.484) said that “The current cost to sequence a patient’s entire genome and the genome of their tumor is down more than 100-fold, but still ranges from $30,000 to $40,000. Prices are still dropping very rapidly; in the next 10 years, it will cost less than $10,000, and it certainly will be more affordable in the next five years,”.

I believe that this dramatically underestimates the pace of technological change. In my mind, researchers and clinicians should be anticipating that, within the decade, a whole human genome will be comparatively priced to a comprehensive blood test. In other words, if there is a compelling medical reason to sequence a genome, cost will likely not be the barrier.

I believe that the impact on the medical community of whole human genome sequencing at a cost comparable to a comprehensive blood test will be profound, and it will raise a host of public policy issues (privacy, security, disclosure, reimbursement, interpretation, counseling, etc.), all important topics for future discussions. My message today is that we don’t have decades to resolve these issues – the technology is nearly upon us.