Fourth Annual DNA Grantees' Workshop

Tuesday, June 24, 2003

AFTERNOON SESSION

Progress of the Genetrack DNA Forensics Chip System
Dan Ehrlich
Biography

MR. COFFMAN: Our next speaker is going to be Dr. Dan Ehrlich. He is the director of the BioMEMS Lab at the Whitehead Institute for Biomedical research. His topic today is going to be the progress of the GeneTrack DNA Forensics Chip System. So I give you Dr. Ehrlich.

DR. EHRLICH: We're committed to bringing you in the near term a really new and expandable system that basically works on your existing principles of operation; that is, using standard DNA kits for electrophoresis and without big hurdles in terms of validation. Basically it employs all of your standard protocols, but it tries to get the advantages which I'll list in a moment that come from the chip. Ehrlich: Slide 1

Scientifically speaking, we think that it's a low-risk project. We know all the principles of operation. We've demonstrated them back in 1998. Now it's just a matter of implementation. Therein lies a long story, and I'm going to give you an update as to where we stand with it.

The main advantages of the chip format over a capillary machine are that the chip can be made extremely fast in comparison, particularly for this forensic type of assay. The chip system has a great deal of potential for expandability once it reaches your lab, for example, integration in that amplification and analysis are all done in a single flow, and also that by changing the chip you can optimize the whole array of different analyses without significantly changing your system or system settings. It also produces a superior result in terms of signal quality compared with a capillary machine.

Our objective, however, is to get it into your labs, and to do it, we took a strategy there of inserting a very high quality result—more or less consistent with the way you are currently doing business—and I'll then talk about some of the expansion we expect from that point.

So today, I'm basically going to give you a tour through the land of engineering. You're probably going to hear more than you wanted to know. It's certainly more than we wanted to know. But at the end of this, you'll know where we are in terms of bringing the system to you.

First, the people who are currently on the project. We've actually had a much larger team during certain periods of the engineering burst, but right now, we're into the final integration and proofing stage. In particular, I want to acknowledge two people who are here today, Nils Goedecke and Brian Mckenna. They are the people who are demonstrating the machine out in the front hall. Ehrlich: Slide 2

Here is a sketch of the basic system. The main point I want to make about this chip system is that it is not a toy. To a large extent, the electrophoresis chip systems that we have seen (e.g., the Agilant system) are very compromised systems. They are very short chips, they don't have any sieving matrix, and they are single color. They don't do what you need them to do. Ehrlich: Slide 3

This is a very serious machine. It has 4 colors, 16 lanes, and enough separation to actually do the separation that you need for the analysis. However, it's as simple as we could make it, with minimum requirements. It was designed with the idea of having the system evolve toward more futuristic applications of STRs (short tandem repeats), for example, and become closer to the field, more portable, and so on. So there was a lot of thought put into this.

There are two units—the analyzer unit, which is a desktop, and a support unit, which will evolve over time, become significantly smaller because of the chip, and be highly customizable to different applications. The optional robot, which actually would be very suitably addressed with some of other elements that we heard about yesterday from work in Louisiana, could be integrated onto your front end so that it could very well service the chip. So that's the scheme of it. It's meant to be flexible and to evolve over time. In particular, this separation into components is an important aspect.

Now, last year when we reported on this, we thought that we were very close. We had the prototype being delivered by the subcontractor and all of the software modules, and we had already proven out the chips. It then turned into the famous engineering point where you're 90-percent finished and 50-percent done. I'll give you some feeling for what happened. Ehrlich: Slide 4

First of all, we were very much enamored with the fiber optic system. It was in the height of the stock market bubble, and everybody knew fiber optics was going to change the world. Well, we had large problems with background. We went through a whole iteration of cycles involving a vendor chain and months of waiting for components. Furthermore, we couldn't solve or get the fluorescence out of the system. We went through two iterations before we came up with a better idea. We launched what was called the Free Space Optics System. After that small delay, I'm glad to say we're back on track. Ehrlich: Slides 5–7

We had to redesign the chip, but fortunately not much else. The entire electronics system—the one moving part in the system—remained in tact and in a very short time, we've been able to prove out the rest of the system. Ehrlich: Slides 8 and 9

Along the way, we were also able to make quite a few chip advances. This is the starting chip that we will introduce. It's a 20-centimeter long, 16-lane chip, so on each of 16 lanes, you get a full multiplex for color detection and single-base resolution out to 550 bases, if you need it. It offers very, very high quality, high resolution, and high separation. This has all been thoroughly verified, so I'll give you a little bit more detail on that.

Also, we were able to develop a demonstration of an ultra-high-speed chip that we believe can extend actually down to something like a 10-second read time. This is based on optimization of an injector. We were able to demonstrate an extremely high signal-to-noise ratio in the system. It is approximately three or four times higher than what you currently have with the ABI (Applied Biosystems). So we're very happy with those developments on the chip.

This is a real sample from Virginia, showing data collected in a single lane. It's a full multiplex. This shows some of the optimization that was done around the chip. Here, we were trying to broaden the loading windows with certain parameters so that it would be minimally sensitive to the remaining salt in the sample. We wanted to make this very tolerant over a broad range of conditions. Ehrlich: Slide 10

The salt concentration in the waste well of the chip varies here. It's just one of many variations. However, the signal does not vary significantly with that particular change in running condition. But here in the center trace, you see (in minutes) the total DNA that's being loaded as a function of this timing parameter, and for the middle condition here we have a very broad and high DNA loading peak. This tells us that this is a nice condition in which to operate the chip. Ehrlich: Slide 11

This actually shows another way we do these optimizations. You can also see this one out on the floor. It's a full sequencing chip that has 384 lanes of sequencing data across it. I'm just illustrating it because it's the same kind of protocol that we use for the 16-lane forensic chips, but basically here it's plotting out quality factors across the plate. Ehrlich: Slide 12

We look for patterns of variation that could be due to geometric effects or problems with lane defects. We can accumulate massive amounts of data and analyze it this way for patterns. This shows the results of the new injector. This is a sequencing run, which is more difficult than an STR run, so we often use it. It shows a traditional application with the traditional injector, good sequencing data, but requires something like an hour to do.

These new injectors—the so-called stacking injectors—compress the DNA on the chip into a very narrow plug and launch it down the chip. With that approach, we figured out that we needed to vary the buffer and conditions in the injector. With that, we were able to dramatically reduce the running times. Ehrlich: Slide 13

When this is extrapolated over to the application of STRs, it predicts the results of a chip several centimeters in length in 10 seconds. Although this is not something that we're going to do in the first iteration, it's very solid proof that you can basically do real-time readout of DNA with this chip system. The GeneTrack system that we are producing will actually have that kind of capability, not on all 16 lanes because of its detector limitation but certainly on a couple of lanes with a modification of the chip.

A lot of work was also done on the software. Brian [Mckenna] was largely responsible for coordinating that. We are trying to produce a machine that really streamlines the whole process, minimizes data reentry, and logs a lot of the critical data that you typically would have to log in a real setting. We've talked with some of you about that and want your input. So if you go talk to Brian some time later in the meeting, that would be great. Ehrlich: Slide 14

We're also concentrating on the data reduction and quality factors of data in the report itself. In all, the whole system is meant to be streamlined with minimal operator intervention.

Here are a couple of slides that show the screens that come up while you're operating the instrument. Basically, it stores all of the settings. It has a very primitive layer, which is the normal operating layer, but underneath it, you can customize things as you need them, for example, you may be changing to a new chip or you may be developing a new assay. Ehrlich: Slide 15

It displays the data for all the lanes in real time. Ehrlich: Slide 16

Here, we are moving into the data-analysis side, and it allows you to look at the raw data. We actually prefer to do this. To some people, it's disturbing. On a professional machine, you never see the raw data; it's always smoothed. But this is the reality of what's actually occurring in the machine. It tells us a lot about optimization, and we look at all the channels. Ehrlich: Slide 17

This is more of the polished end result that you will be using on a regular basis. It displays the data and also such things as confidence factors on the calls in chart format. We expect this to evolve over time as we get feedback from people who are actually using the machine. Ehrlich: Slide 18

This is again more data from a single channel of the 16 channels, looking at ladder data and then looking at an overlay of ladder data to actual sample data. Ehrlich: Slide 19

This is another actual trace, and here you can see the extremely high signal-to-noise ratio, which is 120:1 on the red channel, and when the correct filters are put in, it will be equal on all the lanes, all the channels. Ehrlich: Slide 20

This, I think, is the same data, just blown up to show the extremely good resolution. Ehrlich: Slide 21

That's basically my update. We also are preparing for a tire-kicking exercise. We will ask certain people from State labs to come and look at the system onsite this summer, and after some hands-on experience, we'll have them give us their impressions, including what should be done to refine it. Then we will begin the process of introducing the system to some of the leading labs. Eventually, we'll be looking at the commercialization step. I guess I would say that more optimism surrounds this system than some of the systems that I heard about in the earlier session on commercialization. There is a real market here, and it can be serviced by new companies and new players in the field. Ehrlich: Slide 22

One of the reasons it's relatively easy to address is because the users are very well known. In order to introduce new technology, basically you need to come to this meeting and convince the leaders. It doesn't require a massive marketing operation. You do, however, have to deal with the quality factors because they are extremely important to this community.

But I am actually quite optimistic about the transitions. It puts a burden on those of us in the research-and-development setting to push things further than we had in the past—past the paper that we published in 1998 showing the process and all of this effort in the meantime. So I'm quite optimistic that there is a sufficient market to justify the next steps.

What lies beyond? Well, that's not our focus at the moment, but there are such interesting things as single-use chips. This is a chip of the future. Basically, it's a stamped item that is embossed into plastic. It's a very cheap, single-use type device that's only part-way finished. I mention it just to demonstrate direction.

Beyond that, the next steps focus on the evolution of the machine, that is, shrinking it—which basically requires a modification of the excitation system, which isn't too far off—and becoming ultra-high speed. Ehrlich: Slides 23 and 24

Thank you and please talk to us later.