One-On-One: Jeffrey Schloss

Semiconductor Engineering sat down to talk about genomics, DNA sequencing and semiconductor technologies with Jeffery Schloss, director of the Division of Genome Sciences at the National Human Genome Research Institute (NHGRI). The NHGRI is one of 27 institutes and centers that make up the National Institutes of Health (NIH) in the United States.

SE: The NHGRI is involved in several projects, including the $1,000 genome sequencing technology development program. As I understand it, a DNA sequencer is a lab system that determines the exact sequence of DNA bases. A complete set of DNA is called the genome. What’s the main idea behind this $1,000 project?

Schloss: At the time when we started, the cost to sequence a genome was in the order of $10 million to $50 million. So we had this crazy idea: Could we bring the cost down by many orders of magnitude?

SE: One company, Illumina, claims that it has broken the $1,000 cost-per-genome barrier. Has the industry really achieved the $1,000 target yet?

Schloss: We are getting very close to that, but I don’t think we are there yet. We’ve been supporting what we call large-scale sequencing centers. And they are in the range of about $4,000 per genome if you count all the costs. This is still astounding when you consider that sequencing went from $10 million around 10 or 11 years ago to $4,000 today.

SE: So how can some claim they are doing a $1,000 genome?

Schloss: You can say that if you don’t count the amortization of the machines, and you don’t count the electricity and your labor costs, as well as some of the infrastructure costs for bioinformatics. If you add up those things, our best estimate of cost is really $4,000. If you don’t count all of those costs, then you can obviously state any price you want.

SE: What are the cost trends in DNA sequencing?

Schloss: The path is pretty much downward. But you will see that there is some leveling off. During the time when the cost was dropping so precipitously, there were multiple technologies out there. And there was a lot of competition. But that has settled down a little bit. And we are in a phase now where the cost is sort of leveling off, because there is less competition in the field.

SE: The NHGRI has provided grants to universities and companies to develop new technologies to enable the $1,000 genome, right?

Schloss: That’s one of the things as a federal agency we have been trying to promote. This is the ability for there to be more competition. We support the early research. And then, of course, companies have to support the commercialization of the technology, which is quite expensive. And then they make it in the marketplace or they don’t.

SE: How did you derive the $1,000 figure in your program?

Schloss: There is nothing particularly magical about this $1,000 target. It is true that doctors routinely order imaging tests that cost near or at that amount. The point is that if there is a medical test that returns useful information, and costs about $1,000, doctors will order that for their patients and it’s reimbursed by insurance. That was one of the ideas behind our program. So if you could get the costs down in this range, there would be situations when you would want to use DNA sequencing in medicine.

SE: Does DNA sequencing work?

Schloss: It’s certainty leading towards diagnosis. It is also sometimes leading to therapeutic options. This is more of what we call Mendelian diseases. This is where the DNA sequence is different in one gene in the protein coding region. It actually changes a protein and causes the disease. But for some of those, it’s very hard to figure out which gene has that little flaw in it. People will go through a whole series of genetic tests based on the physician’s best guess. If you can narrow it, you can sometimes tell what’s wrong and it can give you some ideas of possible therapies.

SE: In the future, physicians will order exome sequencing for patients. What is that?

Schloss: It’s sequencing the exomes, or the protein coding regions, of the genome, plus DNA nearby the protein coding regions. That’s about 5% of the overall genome. You can do that now on the order of $500 to $600 per person. For the patients for whom it works, it’s quite informative. But it only works in about 25% of patients. We don’t know exactly why it doesn’t work for the other patients. One obvious guess is that the problem is not within that 5% of the genome. It resides somewhere else in the genome. But that’s big hang-up right now. It’s much harder to interpret the genome sequence outside of those exomes. That’s one of the reasons it’s not done routinely.

SE: What can we learn from Illumina’s new HiSeq X Ten system?

Schloss: This is the one that they are advertising at $1,000, or perhaps less than $1,000 per genome. Because that’s available and cost is coming down, there will be a ton more whole-genome sequencing that will be done over the next year. That’s the data we need to learn how to interpret the whole-genome sequence.

SE: A number of entities provide genomic tests. Is that DNA sequencing?

Schloss: There are different kinds of tests that are called genomic tests. Some of them aren’t actually sequencing per se at all. For example, you can put very short fragments of DNA on an array. There are hundreds and thousands of these sequences on this array. It’s a different process called hybridization. Basically, what it’s doing is looking for a match between a DNA sequence and the unknown sample. That’s the patient’s DNA sample. There are several hundred thousand different locations on this array. If you get a hit at a particular location, then you know that the patient has that sequence in it. You can use that to assess common variance in the human genome. Those kinds of variances have been associated with particular diseases. So that’s what a company called 23andMe is doing.

SE: When will we actually get to the $1,000 genome?

Schloss: I suspect we will be there in the next two to three years.

SE: What are some of the bigger challenges in DNA sequencing?

Schloss: There are a lot of challenges in the analysis of the sequence. You are generating millions of short reads. What we mean by that is the bases are about 100 to 200 nucleotides long. So, in the computer, you have to put them back together. Generally, you map them to a reference genome. But there are a couple of problems there. First, the reference genomes that we have aren’t that good. They are pretty good, but not that good. But we are developing better ones.

SE: What other problems do you see?

Schloss: There are repeated sequences in the human genome. Again, this is limiting the ability to interpret the genome sequence information.

SE: Where does semiconductor technology fit in DNA sequencing?

Schloss: If you are talking about semiconductor technology to improve the computational end, I am not sure that it’s going to solve that problem. We are talking about semiconductor technology to solve the data collection, or how the actual sequencing is done.

SE: DNA sequencers based on nanopore technology are also emerging. How does that work?

Schloss: For this, there is no polymerases molecule. It’s actually just an electronic readout of the DNA as it goes through the pore.

SE: What’s next for nanopore?

Schloss: Oxford Nanopore and other groups are taking the next step of replacing the protein nanopore with hard materials that are nano-fabricated. This could be arrays of many hundred, or a thousand, or even ten thousand, nanopores. These are 2nm nanopores in semiconductor materials.

SE: The NHGRI has funded other futuristic technologies for DNA sequencing, right?

Schloss: In the peer review panels when we are evaluating grant applications, you have people that state: ‘There is no way this thing can work. It defies physics.’ As it turns out, a lot of these things do work. But the fact is we don’t understand the physics very well at these scales. For example, a lot of the modeling you would do with semiconductors doesn’t involve water, salt water or molecules. So part of what people had to do in developing a sequencing technology is to invent the modeling methods to predict how they might design a device or a field-effect transistor.

SE: Why the interest in graphene?

Schloss: When you try to drill a nanopore through a silicon nitride or something like that, the material has a certain amount of thickness. And it’s often 10nm thick or even more than that. Within the DNA molecule, the spacing of the DNA nucleotide is just a nanometer or two. So if you are going to read a signal from that, and your nanopore is the length of several nucleotides, that might obviate your ability to get a signal from the individual nucleotides. So graphene is so thin that it’s possible that you would only have a single nucleotide within the pore. That’s one of the reasons why people look at graphene. The electric field, as it turns out, is a bit more complicated.