Transcripts of the Attorney General's Initiative on DNA Laboratory Backlogs (AGID-LAB) Working Group

March 4, 2002, Meeting

ADVANCEMENTS IN DNA TECHNOLOGY

MR. MORGAN: As part of our continuing effort to explore the causes and the solutions to the DNA backlog, we've really been dividing the issues into two areas. The nontechnology issues as they were discussed by Tim Schellberg, things like expansion of the covered offenders, and that kind of thing. And then the issues that will be addressed by this panel, which are the technology issues.

And I would further divide those into two areas. First, what are the gaps and technology that currently exist, both the gaps and what technology and the gaps available technology that is actually at the crime lab.

And in the second category I put the opportunities. That is what do we need to be looking into for scientific research and development. And to that end, we have four panel members. Two from the state crime lab directors, who are state and crime lab directors. And you'll also be hearing from similar individuals at a later panel.

And what we really want to get from them are what their problems and needs are right now. And from these gentlemen, in particular, what solutions they're applying.

I think you'll find that these two states, Georgia and North Carolina, are putting some very innovative solutions on the table right now with the available technology.

And also try to learn what they see as the needs and opportunities down the road. And then from the science side and from the research side what - find out from them what technology they would like to see put into the state and local crime laboratories, which is an expert systems, and what they think will be available down the road such as chip-based technology.

And the four individuals we'll be hearing from are - first, we'll be hearing from Dan Ehrlich. He is the Director of the Bio-Mem's laboratory at the Whitehead Institute at the Massachusetts Institute of Technology. He is responsible for the NIJ effort in developing a DNA chip system.

His engineering group develops next generation instruments for DNA sequencing, either for forensics or other biomedical applications. He has a very long publication record in lasers, medical electronics, and miniaturized devices from molecular assays.

And I'd like to turn it over to Dr. Ehrlich right now.

MR. EHRLICH: Thank you, John.

I have just prepared a few - very short remarks reminding you of the system that we're developing and updating you on its status, and I know this is a very interactive group, so I'm going to keep it brief.

As many of you know, we have been funded by the NIJ. Thank you, Lisa Forman crew, to develop next a highly specialized chip-based system specifically for forensics. This is chip-based because the data quality factors and the speed of a chip-based system exceed that of a capillary system.

We're also developing specializations to make it easy to use. And I want to give you a quick description of how that turned out in the design. We've been working on this extensively for a period of 14 months now under the NIJ funding.

The key features of the system are that it's able to obtain extremely high speed due to the short analysis channeling of a chip-based system. It is capable of being evolved to a very compact size. I'll tell you where it stands right now in terms of portability. It has extremely high signal quality factors, which we intend to exploit for specific forensic needs such as mixtures and other particular forensic needs.

And it is - we've tried to partition the system to match the needs of the lab. The basis of this system is electrophoretic separation. It uses the standard technology that which is already in wide use in the labs. It, therefore, does not, or should not require new qualification or verification procedures that you have not already established. And that was a key to other chip-based systems. However, they would have required extensive validation. We don't believe that's the case here.

It is being developed initially for bench-top use. And as you'll see actually it's very well suited to a case study application. It can evolve in various directions depending on how the community decides it needs it to go in the future.

Let's me show you where the system is. I do not have this as a Power Point slide. Fortunately it's impossible to get into the format. So what you'll see is there are actually two units here. There's the bench-top unit.

This is the first of three prototypes which the first one has completely been assembled as of this month. The second too will be at the end of this month. You can see there's a bench-top unit. The chip slides in the top door of the bench-top unit. That is connected by fiberoptic link and electronic cables to the support unit, which includes the laser needed for the detector and electronics.

The strategy - this all may very well bust in a very robust format by a vendor which who we subcontracted to use to making mill-stack back equipment. Therefore, it's capable of being moved around in these two units and it would be able to take a lot of the abuse because of the fiber optic link.

So this is the compromise that's been made on portability at this point. It's a bench-top unit but it is very portable. In its current format, the bottom unit can be replaced in the next generation machine and actually largely, nearly completely eliminated using solid, state laser source and summary design of electronics.

So that's how it's currently looking. And I guess I can go back to the Power Point slides. The working element is a 16-lane chip. This is what the chip looks like. It's actually a very long device, and that's because we wanted to get the highest possible data quality out of the forensics assay.

It's possible to make much smaller devises, which optimize at high speed for a type of fast-screening application, but this size is required in order to obtain the very highest quality data which we believe is the first priority for the application.

On this device, I make further other points up here. It can be loaded with a robotic fluid handler. In fact, we use a very inexpensive T-can robot about an $8,000 device which also would be able to set up the PCR reactions. It also can by manually loaded.

Runs in 15 to 20 minutes if data quality is the objective. Can run in short is one to two minutes for a fast-screening application. And that's part of the appeal of the chip-based system.

This shows some of the structure which is in the glass device I just showed - just held up. It shows the 16 channels. Up near the top of the device you can see the press injector structure which is used in which the sample is so-called cross-injected into the channel. That's one of the key features that allows it to outperform a capillary.

This is a short video clip of the sample loading into the channel. What you see is the DNA sample initially brought the orange-colored substance brought down from the sample loading well across the analysis channel. And the analysis channel is horizontal in the slide and the detectors on the left.

You can see this intense concentration of the DNA sample as it is injected into the channel as part of the loading. This is what - a very key feature that allows it to perform at such a high level.

We have run a fair number of samples thanks to the various collaborating labs. These are from the State of Virginia lab. We are able to have run all of the current and even the experimental multiplexes in a single channel. So we have 16 channels. Each capable of handling full multiplexes.

I'm summarizing now. I think I've touched on many of these questions. It will do a standard multiplex in about 15 minutes with very, very high data quality, sixteen channels. There are issues which remain as to how to integrate it into a standard forensics lab usage.

And this we will begin our collaborations, which are scheduled for the two Florida labs, the Virginia lab, and the State of Massachusetts. And these will be very imminently and certainly in the next quarter and will begin to figure out how to integrate these into the lab practice and to find out if we require any changes in the lab practice, but I don't believe we will. We will figure out how to connect this with automation.

One question we got is: Will the chips be disposable not in the first iteration? There's no requirement to do so anymore than there is in a capillary system, but certainly it's possible to evolve to a disposable, probably plastic-based chip.

Will the system be portable? Well, initially it will be portable in the sense I've just shown is it's going to be a very robust system which could be transported without loss of alignment and so it should be transportable in that sense. It has a potential to go further.

Software. We're going to hear from Mark next. We'll be integrating with Mark Perlin's expert systems software.

When will it be ready? Well, it will go into its first beta test in the next quarter. Over the course of a year, we will be interacting strongly with the state labs. And state personnel will be coming into our facility and running the prototypes. And so sometime we'll have the detailed schedule as to what happens after that is in discussion at the moment.

I've already mentioned who will do the beta data. And we welcome additional input from people as they may have it. So just contact me or Lisa.

And how much will it cost? We always get that question. I don't know. There is not a detailed plan to bring it out into the labs at this point, but it's certainly designed as an inexpensive system. All the choices were made such because we know that this community is capital adverse. So that's my summary.

MS. HART: Could I just ask one quick question and part of it is a conceptual one on my point of view. Obviously not coming from a scientific background here, but conceptually how exactly is this to be used?

In other words you're talking about let's suppose somebody is taking a rape kit or let's suppose somebody is at a crime scene, how would this - when would this chip come into play and who would be using it and how does it get to the lab? I mean, I don't know - it's beyond me.

MR. EHRLICH: We saw this as taking and probably being inserted in a sequential fashion into the crime lab. The first thing it would do would be simply to replace current analysis systems which are not as expert of doing the application or as well as streamlined in doing the data logging and all of the overhead related to casework. So this would go directly into the state crime labs and be used as to streamline the current practice.

It does have potential to be used in other applications, including closer to the crime scene probably still with an expert operator but this - the first stage of this would be to go directly into the labs to replace and streamline current practice.

MS. HART: But the idea of having a portable system is ultimately so that it could go outside the lab?

MR. EHRLICH: In principle it has some considerable ability to do that, because it's a small machine and it's been - all aspects of it are small, but in principle can have a little power consumption and those kinds of things.

There is some additional lab practice, which also would have to be brought to the crime scene, for example, which would include the sample preparation steps and so on. So that is a bit more complicated technically as well as having very social issues.

MS. HART: Thank you.

MR. MORGAN: Our next speaker will be Mark Perlin. He is the President of Cybergenetics. He, in general, develops biomedical information and automation technologies.

Some of you are certainly familiar with one of his previous products which was the - he directed the development of the TrueAllele automated data scoring software, among other past projects. He is the CEO and founder of Cybergenetics, hold adjunct faculty appointments, and computer science at Carnegie Mellon and in Human Genetics at the University of Pittsburgh and will be discussing expert system software.

Dr. Perlin?

MR. PERLIN: Thank you.

So they'll be talking about the automatic meat interpretation of various forensic DNA data, expert system technologies for reducing the backlog.

Just a quick note. What is this data for the two or three of you who may not see it every day? The first on type of data is when you have one person or a pristine profile, you get one or two peaks representing the two alleles.

And then in casework across a number of loci you may end up with many contributors. For example, here there were four peaks corresponding to four alleles. It can get more complex from there. So this is what we're talking about data at one locus or one genotype.

In generating STR data, over the last ten years we've seen quite a lot of technology and automation. We've gone from a cottage industry to having automated sample preparation and extraction, automated thermocycling, and automated sequencers.

In the U.S., as we start the process of interpreting all of this STR data, though, we end up with lots of people who may use computers as part of the tasks, but spend a lot of time of looking at data and then chatting with each other and discussing it and looking at every peak.

So this has led to an interpretation bottleneck where the data generation, which is done by machine, largely overwhelms the human review. So what's needed is some kind of computer automation that provides the quality and assurance and the integrity of our databases and can be used for casework and mixtures. And the key goals here are having no error, having very high input, and having no staff or small staff.

(Laughter.)

MR. PERLIN: So TrueAllele technology is something that works in eliminating the NDIS review bottleneck. The idea is that the information, the raw data and the original data comes from any gell-based sequencer or capillary instrument.

The computer automatically without human intervention, running on most any computer, Macintosh, Windows, UNIX, does all the necessary steps, color separation, and image processing and tracking, single analysis, ladders and so on, down through quality checking and reporting.

And the output of this is quality assured profiles, which can be put onto a database. I can't go into the thousand of things that the system does and the dozens of things we had every week. So I thought I would just tell you about one thing, a data quality rule.

The system has dozens of rules, which if used, is to check every genotype. This is an example of one rule. They're all, of course, user-customizable to whatever the thresholds are, whatever a user wants, whatever the groups' SOPS might be.

This is a third peak rule. And it's designed to be in accordance with a state standard operating procedure for how they would want a third peak to be recognized. The idea of the system is that if you passed these 20 or 30 rules, then the data generally, and you'll see stats in just a second, have no problems and don't require extensive or any human review.

However, the 10 or 15 percent that might be flagged, the computer has diagnosed it and it tells the user what the issues might be and then we suggest that a person looks at it and make a decision.

So the TrueAllele process is something that packages up how this software is used. And looking at what we do we find that the software itself is barely 5 percent of the lab's using this.

Our concept of it - this is a desktop system. This is a photo I took from our office the other day. This is how we office data. Essentially it's a factory of computers that just sit there happily processing data. And a person is involved in another room or in the same room. And what the person does is, like on a production line, checks what the systems check have been.

So the computer does all these checking and quality assurance and calling and whatever needs to be done. And then a person monitors that the computer has a sanity check and is working the system properly. Then at the very end that 10 or 15 percent of data can be looked at.

The three-foot of this of just looking at the data that needs to be looked at in our validation studies comes out at about 100 genotypes measured as one locus of one sample per minute to cross several platforms. The validation studies were done. Once we had the original data, the concept is the expert system did most everything. Then the computer would make a decision of accept, reject, or edit.

In production, a person would only look at edit. However, in validation, we would review all the data and the person would agree or disagree by accepting, rejecting, or editing.

The validation results from Palm Beach, FDLE, New York State, Virginia. We continue to work on various studies that generally show that 85 percent of the data across these platforms, FM Beyer 3,700, 3,310 and so on, do not really need to be looked at or looked at that carefully. And that there's basically no error.

When we do get anything at all one, even one in 10,000, one in a 1,000 issues coming up, we then go back and retune the system to that lab's specific needs so that in the end there is no error, but we maintain the efficiency. So the result is the computer can do about 85 percent of the data without people looking at it and the designations are correct, which may help eliminate that backlog.

With mixture interpretation, some of you may know we've been doing quite a bit of this on automated casework. In doing the analysis and the reporting, we've built about, I guess, the tenth system, prototype system, in the last three years on this. And I'll give you a quick note about what it's doing.

There's a lot of complex math. This is a simple bit that was published in general forensic sciences in November. And there's I don't know how many man-years of effort, but have gone on into making this mathematics into something very robust for general casework. I won't go into mathematical or statistical details today.

The first thing that we're validating is a rape kit process. And, again, I say it's a process because 95 percent of the work in how we design these things is thinking about how do we get from the original data all the way through to what the report will be in a way that might be useful with audit trails, ports, and so on.

So the original sequence of data comes in. And then if it's batched data it goes to TrueAllele. And all we do here are the quality assurance tests. Nobody actually looks at the designations at this point. We're just making sure the data is of high quality and looking at the controls.

The output of this initial batch processing is a quantitative peak database. And every peak has its quantitation (sic) attached to it, and that's the input that's used for the automated mixture analysis. And then the system does the interpretation, the reportings, generates figures, statistics and so on.

The first run-through we did, we'll talk about in a second, on a per-case basis, there was maybe five minutes of human time spent in the generation of quantitative databases from the original sequencer data and about one minute of human time just looking over what the case did. It's still too slow.

The LMA rape kit validation, we started off - we actually have data from a number of groups. And as we move the technology, we start doing studies.

With Illinois data, we looked at 25 cases. These are non-suspect cases. We looked at the mixture plus the victim and then infer the suspect. The linear mix analysis that's reported in the literature is the starting point for that. And then the computer goes off and does a lot of sophisticated statistics for the other 99 percent of it.

So what you end up with - I'm intentionally showing you a case that's somewhat ambiguous, but let me just walk you through it. The computer automatically generates thousands of pictures like this. Here is the known victim profile. This is D-18. This is the observed mixture data. This is - on the third pane, a model of what it might be with the mixtures computerized and the different contributors inferring this genotype.

In most loci, this is a 25 percent unknown suspect. Most cases, the third panel completely matches the second panel. That is the model matches the data. However, in real data, you often have ambiguity and the result is the computer will compute - it may say that initial answer of 1519 has a 60 percent probability, but it goes through and reports the probabilities of any ambiguous loci telling you what the genotypes are. It can do it at the full genotype level.

It's most useful for CODIS applications to report at the individual locus level. Most of the time it gives you a unique answer, but when it doesn't, it tells you the extent of its uncertainty. The results are objective, complete. It looks at basically everything. It's accurate.

We're working on efficiency, five minutes per case is okay, but there may be more cases coming along. It's fully automated.

What I found most striking in this first validation is that it occurred to be only after the fact when I sat down with the forensic scientists, but I've never seen it peak from the data. It was just all done by computer from the peak database. And it generates all these reports such as these statistics and the various useful figures.

So where we're going with this is looking not at how to infer profiles, but how in this case if you had,say, two crime stains, each in yellow, with a low level contributor, how the computer can pull out the exact probabilities of each genotype. And then from that, of course, using CODIS pull out who the perpetrator might be.

I hope in two months or so we'll have some testing on new stuff for multi-scene interpretation. Here, again, yellow may represent a bit of contributor for multiple scenes. And the goal is to pull out relatively unique profiles that can be used with a goal of one day having computers doing continuous DNA surveillance on quantitative peak-based data of the type that you see in your lab across the country giving you the ability to identify people, apprehend them, but also to convict them, because all of the statistics are done up front and you know the exact certainties with which these profiles are unique.

I'd like to thank our collaborators who have provided us with the data, the Florida, Virginia, and New York labs, the people in our group, NIJ, who has been very generous with their support, and our host, the U.S. Department of Justice.

Thank you.

MR. MORGAN: That's great stuff. Thank you.

George, are you cued up first or is Mark?

MR. HERRIN: Mark is.

MR. MORGAN: Mark is, okay.

The next person on our panel is from North Carolina. It's Mark Nelson, the special agent in charge of the molecular genetic section of the North Carolina State Bureau of Investigation, a charter member of SWGDAM, and an inspection team captain for ASCLD-LAB and an NFSTC. And he is also serving as program manager of the National DNA Audit Program with NFSTC.

Mark?

MR. NELSON: I'm going to give you a little bit different perspective today from a state laboratory level. And I'm going to do this by taking you on a historical trip first, because I think it's important that we look at where we were as to where we're going.

Going back to the old days, we started our laboratory in 1989. We were using RFLP Technologies. And I start today talking about databasing first and then switch to casework.

We prioritized our samples. We did sex offenders and murderers first. And we got pretty far along that process and the STR Revolution came along.

And the point that I'm trying to make with this slide is we have to be very careful about changing technologies, because they're going to directly impact our backlogs. So we went from having a database that was partially complete to having a database we had to start all over again. So it's very important that we be careful of our technologies changes.

The reality for dealing with these types of things, of course, is the only people who can get blood out of a stone is the IRS, and resources are very limited. So with that in mind, we decided to set forth two types of strategies. A short-term, which we would outsource our samples and then simultaneously work toward a long-term goal of building the infrastructure within our own laboratory and using new technology to handle the samples ourselves.

So our first strategy, we did outsource samples. We did get NIJ assistance and were very greatest to that. They did come to our rescue. In one year, we outsourced 8,500 samples, and in the next 14,000 convicted offender samples.

And this year George and ourselves, I believe, are the only two laboratories that sought NIJ assistance to do our own samples in-house. And one of the reasons for that is we were using our long-term strategy in developing new technologies to help us in this.

But the other thing is that we found that 30 percent of the effort of outsourcing we had to do in the front end and the back end. And it was - just made good sense to go ahead and do it all. Why not do the other 70 percent in-house.

It gives you some idea where we were last year. We had 47 percent - I'm sorry, two years ago, of our backlog untested. This year we had - or at the end of 2001 we had 9 percent. And of those 9 percent, two-thirds of those are now in CODIS and the one-third will be in CODIS by the end of this month.

Of course, one of the nice things about the NIJ grants was that we do a lot of unsolved cases and we obtained the 17 percent roughly hit to those unsolved cases. And the numbers are actually outdated, because last week we got a hit to another one of these cases.

So some thoughts on outsourcing. The backlog is down to the point where we could handle it ourselves. It gave us the time to validate the automated technology that we needed to handle these things in-house.

The other thing that we knew is that federal funds aren't going to last forever. It would be a new baby that politicians will want to kiss one day, and the funding will eventually dry up. And as long as we get federal funds, we may not get our state legislature to give us the funds that we need on a continuing long-term basis.

So with that in mind, we decided to pursue simultaneously our long-term strategy, which is using new technology. In '98, we put robots in our labs. Unfortunately that particular robot didn't do what it was advertised to do. But in 2000, we did put kides (phonetic)and robots in and they did perform. This is the exact robot that we used in our laboratory.

The beauty of it is that it will handle 96 samples simultaneous. It takes three hours to do a run. And we're routinely making two runs a day, or we're getting about 200 samples through our laboratory a day. And these are convicted offender samples now, not case samples.

We coupled that with another robot which is an automated pipetting station. And if we actually got this thing to work with one micro liter in volumes. And it does all of our amplification setup stages.

And we actually had individuals in our lab that were getting repetitive motion disorders from doing all of this pipetting states. And we have now eliminated that by getting machines to do that grunt work for us. And it does it much, much quicker.

Well, the next bottleneck that we ran into is we were using blood on paper. And the analysts were hand-punching these samples. And there, again, they were complaining of sore hands. They're doing so many of these things. And we were concerned about sample mix-ups.

So we evaluated some automatic hole punchers, and we found one that had a bar code reader on it. It would automatically generate the sample order, if you will, in this 96-wheel plate. It will not allow you to put the same sample in twice.

And it's an exportable file that we can carry through the entire lens system through the whole process. And now it takes twenty minutes to punch a plate instead of two hours. This is the unit that we used. And you can see the bar code reader on the top there.

The next bottleneck that we had was we were using a gel-based system. We were running ten gels per day. It's actually ten times two because we used two different amplifications reactions. And we were processing 250 samples a day.

Well, that was done during the day. We want something that will run at night so that we can during the day extract and get these robots to prepare the samples and then run them at night and have them ready for us the next morning. So we now purchased two capillary electrophoresis units so they'll do that for us.

The next bottleneck: Reviewing the data. And this was a real bottleneck. So we are looking now towards expert systems, which Mark just talked about, to help us with these efforts.

And hear my thoughts. And, again, now say that these are my thoughts that I don't represent those in my agency. I would recommend that NIJ use the $1.5 million in key-c funds that they have to purchase expert systems and put hem in the databasing laboratories. That's where they're needed in the labs that do databasing, because we deal with large quantities of samples and batch formats.

I would further recommend that that be part of CODIS software for those databasing labs so that when we get a new CODIS upgrade, if there's a new expert system upgrade, we automatically get that to us through the CGIS WAN. I think that would be a wonderful thing that NIJ could do for the community.

Let me switch now and talk a little bit about casework. You just heard about chips. Certainly they're going to give us a faster turnaround time on those rush cases. I really like the idea of doing it in the laboratory where they'll bring us a rush case, and they want the answers like right now.

We keep an analyst in the laboratory all night. We can get them a result by 8:00 o'clock the next morning. What I'd love to be able to do is do this in twenty or thirty minutes while the officers are waiting there. That would be wonderful for us and then eventually take them to the crime scene.

How should we use the chips? I get terrified when I hear about we're going to give officers DNA chips. I think that what we need to do is make sure that it's a qualified DNA analysts that's doing the testing. You have to be in order to put it in CODIS and do a search against CODIS.

And what I would really foresee is we have agents or scientists or police officer scientists that go to crime scenes and utilize this technology to develop the suspect profile, run it against CODIS. If we get a hit, we can then get a search warrant, find the guy, serve the warrant, do your sample, and then issue a lab report and attach it to an arrest warrant and he goes to jail. The case is over with and done.

I've given some thought to chips and how we could use them at the scenes. And one of the first things we talked about obviously is speed. But if we have to look at the questions versus the lung samples, because quite frankly from the time it would take me to get from my lab to a scene, is about the same time it would take for an officer to get from the scene back to my laboratory with a bench-top type unit.

So the real speed lies when you get a hit, and you can then process that suspect standard right there at the scene. That saves a tremendous amount of time. But I think the real benefit to us is it's going to lower the risk of contaminating of the evidence.

You'll have a scientist on the scene which is going to increase the credibility of your evidence in court. And finally when you go into these scenes, and I've been on over 200 murder scenes in my career, there's blood all over the place. How do you select which samples to collect?

And by being able to screen right there on the scene it will allow us to get those critical samples when you only have to do ten of them before we find the suspect's blood at a murder scene. And I think that's where the real benefit is going to lie.

So where do we go from here? Again, I'm full of ideas and I'm giving them to you. I think we need to switch the funding emphasis from outsourcing to building the infrastructure in the state and local crime laboratories so they can handle their own work.

And to do this we would need immediate funding for equipment, supplies, space, and most importantly people. Because with casework, it has to be people-based. We just don't have the resources to do all these cases if we don't have the people.

We were not able to participate in the backlog reduction - casework backlog reduction because we lack the people to even process the front end and the back end of these things and enter them into CODIS. And I think that's a real shame.

So how can we do it? If you can't fund the people, maybe you can provide incentives to states to beef up their crime labs by tying it to hopefully a carrot instead of a stick, i.e., some ideas.

If you want federal highway funds, upgrade your crime labs. If you have more cops under the cops program, upgrade your crime lab. Put people in there. That may be the kind of incentive that we need because 98 percent of the crime scene work is done at the state and local level. It's not done at the federal level.

And the way that you guys can impact the casework is by providing us with the resources or the incentives to get our people to give us the resources so we can do a job we're required to do.

So priority one. Let's make sure we stabilize the type of testing we're using. We're talking about backlog reductions. Ninety-five plus percent of all of the cases that are out there that need to be tested can be done with STRs.

Sure the snips and the mitochondrial and lyochromes and stuff is very important but only to a certain percentage of the cases. So let's look at the majority of the problem first and concentrate our efforts there. I'm not saying we should cut down on the research end of it. But if you really want to put your bang where your bucks are, you're going to have to go to working unsolved cases with the STRs so we can compare it to the database.

My laboratory was in the process of setting up a minor lab. We had to withdraw that because we felt it was more important to be able to address 150 unsolved cases in a year instead of only ten mito (sic) cases.

So my final thoughts. Outsourcing is a short-term fix, and we have to use the resources available now through technologies to increase our capabilities, and we need to build up the infrastructure in our crime labs.

MR. MORGAN: Thank you very much.

There was things that you said that really stuck in my mind as a former state legislature and that is when you mention that 98 percent of the crime work it occurs at the state and local level. And we do need to keep in mind that this is really an area where the federal government is here to provide a helping hand to the people who are really having the problem. And that is the state and local crime labs.

Our final presenter on the panel is George Herrin. He is at the Georgia Bureau of Investigation Laboratory, and has made several stops around, including Rice University, Texas A&M, Cellmark Diagnostics.

He joined the Georgia Bureau of Investigation 13 years ago and was responsible for the establishment of their DNA testing program. He is now assistant deputy director of the Division of Forensic Science there and has shared responsibility for the establishment of their quality system accredited to both ASCLD-LAB and ISO standards.

Some of their statistics to the end of February 2002, they've done a total of 81 forensic case associations. Ninety-five links to specific offenders and 224 aids to investigations, which I think stacks up pretty well across the country.

George, take it away.

MR. HERRIN: What I would like to do with a slightly different twist, if I can get this Power Point to go forth. I want to give you a little bit of a background about what DNA testing at the GBI laboratory system is like.

We have seven laboratories in our system and two of those laboratories have DNA testing in them. But out of those seven laboratories, we only have 15 staff members involved in DNA testing. And you'll see that that's a fairly low number compared to the number of amount of output that we're doing.

And I broke it down here. And Mark and I would really like you to pay attention on this one slide here, the fact that seven scientists are involved in reviewing offender data. And I think this goes back to what Mark Perlin was saying is that we need expert systems to help in this.

Now, in calendar year 2001, we entered 31,793 samples into CODIS. We're an all felons state. We do all felons as they go into or leave the system. A prison system who have stayed in a state-paid-for bed. So if a person gets out on probation or something, we don't get to collect those samples.

It's also nearly over a little over 3,700 forensic biology service examinations. And this includes everything from blood examinations to DNA typing to semen identification. And as you can see, we did just a little bit less then a 1,000 DNA-type examinations, and 164 of those were no-suspect cases.

This has been an ongoing project with me since I was involved directly with the DNA program and that was that we would put a lot of emphasis on doing the no-suspect cases, because the CODIS database is useless without the no-suspect cases being worked. You're just spending a lot of money and a lot of effort to produce a database which is trivial and not going to help you at all.

Now, where are we standing right now? This is active cases that we have in a backlog. We have a little bit over 11,000 services which are incomplete as of last week. Most of those are not in casework. We're fairly well caught up in casework. Most of those are in the offender samples.

We're working very, very fast to get the offender samples done, but we haven't gotten there yet. We hope to be at a zero backlog in offender samples by the end of this year, and we're going to do that using the technology that I'm going to show you here in just a new minutes.

Now, for the equipment that we're using for DNA testing, we also went with a robotic system. We have two of the kides and bio-robot 3000, which are the customizable units from Kiajon (phonetic). Each of those systems can process about 350 samples a day.

So we have the capability of processing nearly or extracting DNA from about 700 samples per day. We have two ABI 3100 CE systems. This is the 16-capillary system and then 1 ABI 310 system which is used for a backup or for problem samples.

And then we have three of the 9600 or 96-well thermocyclers involved in just doing the convicted offender samples. And in all of those cases we did - you see we have multiple instruments of each type and we did this as an effort to back up so that we would have redundancy. If any one instrument fails, our program does not shutdown.

For casework, we have five of the 9600 thermo cyclers and then seven of 310s. We're hoping in the next year or so to get another 3100 to support casework samples. That's going to depend a little bit on funding.

Now, the average caseload, as I said, we only have 15 people or staff members statewide devoted to DNA testing. And as you can see, that means that there's a lot of casework being done per scientist. 314 services per year or 26 per month on average. That's a lot. This involves everything from screening of the evidence to locate the stains, to the actual testing of the items for DNA and there were 115 DNA services a year or 9.5 per month per scientist. So it's an extremely fast work pace in our laboratory.

Offender samples, we only have one scientists currently tasked full-time for processing the offender samples, because of the robotics, once scientist and one technician. So that one scientist is tasked with doing nearly 16,000 samples per year or 1,324 a month. So they're really moving.

Now, the reasons for the high productivity is that we've gone to extreme automation of extraction procedures. All of the extractions are done with robot. We're using a buckle-swab collection. You just break the swab off and put it in the test tube and the robot takes it from there and you don't have to touch it again.

Standardization of procedures. Our procedures are exactly the same in each one of our laboratory sites. There's absolutely no variation whatsoever. This makes for a lot of good things to happen in which I'll talk about in a minute.

Computerized record storage: LIMS. We are 99 percent paperless work environment in our laboratory. We keep no paper records whatsoever.

Accreditation. A lot of people don't like accreditation, but I think it's a very good thing. And then the professionalism of the staff obviously.

Now, what is the impact of having a LIMS or a laboratory information management system. We can keep an electronic chain of custody. We don't have to continuously sign chain of custody documents. Every piece of evidence, including the offender samples is bar coded.

Standardized reporting for casework samples. You don't have to worry about how to report a statement. It's already in the computer for you.

Electronic signatures on the reports, we don't have to manually sign reports anymore.

Internet distribution of reports. We have a website that all of our reports are distribute through. We don't mail reports anymore.

And ability to batch up and date the status of services in a batch of cases. If you want to do them, you know, 90 to 100 services at a time, and digital photography and input of all analytical results. All of the results are on the LIMS systems so they can be reviewed right there in one place for a case. You don't have to be looking through your file cabinet for paper to review a case.

Now, why standardized? Training times are decreased and made for effective. Procedures are easier to troubleshoot when problems arise. And data review is streamlines, not streamlined enough, but it is streamlined.

Now, the value of accreditation is we have recognition in the community. We have fewer challenges in court, and it does help foster the atmosphere of continual improvement and process optimization. There's not a month that goes by that we don't improve our processes in some way.

What are the bottlenecks? The labor-intensive manual methods for casework sample DNA extraction. There's been a lot of emphasis over the last eight years and then proving the typing systems that we used, everything from the chips to the capillary systems. There's been absolutely or very little emphasis on doing the front end of this whole process, which is getting the DNA out of the samples. That's where we need to be putting our focus.

The evidence screening prior to DNA analysis. You can't do DNA on everything that comes into the laboratory. There's not a laboratory or a government in the world that could afford to do that. So you have to be able to screen the evidence. There needs to be a better way to screen the evidence.

Data interpretation and review for offender samples. I don't think that I agree with Mark Nelson here that data review on offender samples could be automated. And it certainly would be much better. I'm a little bit cautious. I think Mark Perlin's program of TrueAlleles will be helpful, but there's always going to have to be that human touch, I think, in the casework stuff or at least for awhile.

Obvious problems of having a big backlog. You have unworked cases. The victims and the suspects are in a state of limbo. Additional crimes are perpetrated because serial offenders haven't been identified. We had a case in point of that in Georgia where we had a serial offender who committed 2-something rapes because he wasn't in the database.

Delays to the criminal justice systems and then wasted investigative resources. I can't tell you how many hours investigators waste just because they're going down the wrong path. What are the intangible results of having a large backlog. You have a decrease morale of your staff.

Continuous request for rush analysis. I've got to have this tomorrow.

Increased staff turnover. When the morale goes down, the staff leaves. The staff burns out. And evenly if they stay, they don't produce very much.

And then the new technologies, and that's what this is all about, the new technologies don't get implemented, because you've got such a crushing load of casework, you don't have time to implement anything.

Now, the current high-through put technology that we're using and is the robotic extraction of the offender samples. Multi-capillary array or high-density jail-based typing systems, the 96 - the 90 - what are the 3700s that you're suing in Florida?

Commercial kits that enable simultaneous termination of all CODIS site. We've just switched to the single kit from Cook & Elmore or ABI, I believe, so that we can get one amplification. And then software that can be semi-automate through the use of macro programs. And that's the genotype or the - what's the one FM-bio that - star base.

And all of those things help, but they're not the answer. What could really help? DNA extraction procedures of casework samples using robotics. This really needs to be a focus. We need to look at being able to extract DNA from sexual assault cases or even just blood cases robotically rather than by manual methods.

Software interpretation and comparison of offender profiles prior to or forced submission to CODIS. And by this I mean that you run it through two separate computer programs and then compare the results from the two programs. Those that have complete concordance get automatically uploaded. Those that you have a discorded result or a problem, then you manually review those. That would eliminate probably 95 percent of the manual review that's done now.

Contract employees off-site or internal to help in processing until backlogs have been eliminated. One of the strategies that we took in handling the backlog that we knew that we were going to develop with the offender samples was hiring temporary employees to do the front-end work of entering the samples into the database or into the computer.

And this is what will vary the level of temporary staffing based on the volume of samples that we have to deal with.

And then finally screening of sexual assault cased slides for presence of spermatozoa using robotics or instrumentation.

This is one thing that takes a lot of time that not many people think about. For an averaged analyst to screen a sperm slide - a semen slide for sperm might take an hour, if there's only one or two sperm on that slide. If you could do that with a cell-sorter-type of instrument then you would really speed that analysis up.

And that's the end of my talk.

MR. MORGAN: Thank you very much.

We'd like to open up the questions for the panel.

MS. CROUSE: George, I'm just curious. I know that most of the people in the forensic laboratories are familiar with the users group internet system that use to kind of just get out and talk about some things that are happening in the laboratory.

And one of the recent questions was: How many samples per case do you have? And I'm just curious how the State of Georgia, because there was an answer from you guys, in which you do three samples per case, and how did you do that?

MR. HERRIN: Well, what we try to do is we try to really thoroughly screen the case prior to sending it to DNA and pick the one sample, the one unknown sample that's going to be the most probative to answering the question in the case. Was the suspect involved in this particular crime? Or was he possibly involved in this particular crime? And if not, then we go to the next sample, you know, if we don't get an answer from that. But we always start with not more than three to four samples per case, and, you know, two of those being know samples.

MS. CROUSE: So if you do a case and you have a sample and it does not match the suspect but they might have another suspect or whatever -

MR. HERRIN: We'll continue to screen suspects as long as they bring them in.

MS. CROUSE: But you'll take them all the way through DNA and then start over and then take them all the way through DNA?

MR. HERRIN: Uh-huh.

MR. MORGAN: More questions?

MR. FERRARA: And this is directed to Mark and Dan. Has there been any thought given to marrying TrueAllele to gene trace down the road or -

MR. NELSON: Yes.

MR. EHRLICH: Yes, we're planning to do that. We've already been - had a lot of discussions and meetings.

MR. MORGAN: Barry?

MR. SCHECK: Mark, you mentioned something about customizing some people's SOP and then editing.

MR. NELSON: Yes.

MR. SCHECK: Does your program provide for making a record of all those edits because I can assure you -

MR. NELSON: Absolutely.

MR. SCHECK: - people are going to ask for them.

MR. NELSON: The software keeps track of a lot of audit trails. And one of the ones that we are actually testing out now for New York and Virginia the program keeps track of all the edits. Now, what we're doing is taking many of the audit trails the program kept inside in generating reports. So that when something is edited, you know, what was edited, who edited it and what it was, what it was changed to and so on.

MR. SCHECK: So if somebody were to go back, you'd be able to see -

MR. NELSON: What the change was -

MR. SCHECK: - they were -

MR. NELSON: - and who made it.

MR. SCHECK: - and you'd be able to produce a whole electronic record of that from beginning to end?

MR. NELSON: Exactly. It is being generated.

MR. MORGAN: Sue?

MS. NARVESON: This is for Dan and the chip technology issues. Just to clarify in my own mind and perhaps to help Sarah with the question that she asked.

When we talk about chip technology and total port ability that we could actually take to a crime scene, what you're working on right now is the analysis portion of the ultimate, which I see as being able to take a blood sample from a crime scene or a biological sample from a crime scene, put into a small device and actually have the extraction PCR set up and amplification and analysis done in one compact unit.

MR. EHRLICH: Well, we're not explicitly working on the two front-end steps you mentioned, which were the extraction and amplification. In principle, that can go on chip in some future version. There are some technical issues related to the optimizing the chemistry so that it doesn't interfere, the amplification doesn't interfere with the assay.

But the scientific is working those types of issues so it's not out of the question in the future. Mark made points, which I agree with a lot, about the necessity of having an expert operator in the loop.

MS. HART: Mark, you mentioned some questions about staffing issues. You raised some staffing issues here. Even if there were unlimited funds, would you have sufficient pool of competent people to be able to hire to a level that you would need to hire to do what you wanted to do? And the second part, you had also talked about the idea of having analyst doing crime scene work. Would there be a sufficient pool of qualified people to move into that area and would it be cost-effective given how much they might be required to be paid?

MR. NELSON: The answer to both questions is: No. There's not enough qualified experienced people out there. And what we get is a lot of applicants from some forensic science programs, they know a little bit about everything and nothing about anything in particular.

I would really like to see an emphasis on forensic biology track where they actually come out and they've got all the course work necessary to have the experience they're use to the platforms.

And Paul has an excellent program with VCU where he has a forensic science foundation right there in his lab and he trains them on the platforms. And essentially they go through his training program. And I think what we're going to have to do is we're going to literally going to have to tie the university training programs to a crime laboratory in order to get that qualified pool of applicants.

And as far as people going out to scenes, no, we don't have enough people qualified to do that either. In most police departments, unfortunately the guy that has to collect the evidence and get dirty is the lowest guy on the totem pole, because the best investigators want to sit down in their nice suits and talk to people. They don't want to get their hands dirty and crawl around in the blood and the gore.

And we got to get scientists out there to collect these samples and be able to know enough about what the crime lab can do with these samples so that they're collecting the right samples for us.

MS. HART: Given the fact that you think the detectives are concerned about going down there and getting dirty in the gore, I guess as you described it, is it realistic to think that scientists are also going to want to do that too.

MR. NELSON: Well, in my laboratory, all my scientists are also police officers. And all of us get down in the blood and the gore at crime scenes and do what we have to do.

MS. HART: Okay.

MR. NELSON: So, yes, you can do that. You have do find a special kind of person, but, yeah.

MR. HERRIN: But if could speak to that just a second. In our laboratory, all the crime scene work is done by investigative agents. The scientists do not go to the crime scene, except in very rare circumstances.

I think that there is a fairly good qualified pool of applicants because we have to turn down about 20 applicants for every position that we have. I think that one thing that could be done, though, that would be NIJ sponsorship of fellowships or something in the laboratories, I think that would be a very good program that would allow us to train people as they're going through college, you know, getting their undergraduate degree to get, you know, have a trained - a pool of resources to hire when they get out of college.

MR. MORGAN: Dean?

MR. GIALAMAS: I was just going to put in a comment about the crime scene. At LA County Sheriff's Department all of our criminalists do respond at the crime scenes. That's one of their functional duties as scientist. We do respond out to an average of about 10 to 15 crime scenes a week, so we don't attend all major crime events, but most that occur withing the county.

And our numbers fail in comparison to the numbers you've seen. I mean, our DNA analyst can only turn out about three cases a month, because of the fact that they have added responsibilities between court training, crime scenes, and now doing DNA casework. So their time is split.

And I think what you will see across the country is variations from one extreme to the other. From my extreme, where we're attending many crime scenes. To the other extreme where you have DNA analysts who are just in the laboratory and are not required to respond out to crime scenes. And that has a great effect on the overall service, but I want to echo what we've heard, and that is that I can tell you from the crime scenes that are criminalists have attended to we get a much higher quality level of evidence that comes out of those crime scenes. And because they're involved in the collection and being there, they can be more selective in choosing the stains that will yield some better probative results.

MR. MORGAN: Maureen Casey?

MS. CASEY: I just wanted to make a comment as we're talking about the utilization of resources and backlogs and the best use of the scientists. And I think it goes to the issues we were speaking about earlier about law enforcement education.

I think part of what we need to look at is that, in addition, to educating the law enforcement professionals, the bosses about the uses of DNA that if we educate the line officers and the evidence collection technicians and those folks, not only about how to collect but what to collect what's most probative, then we get a balance on the best uses of the resources that are out there.

So I think that while in the idea if we could have a criminalist or a scientist at ever crime scene that'll be great. But as the expectation of collecting DNA at burglaries, as well as sexual assaults and homicides, and you know, as that all grows, we're going to need to use the resources that we already have, which are the cops and the crime scene analysts that are already responding to every single crime scene. So I think we need to bear that in mind as we talk about this.

MS. HART: Sir?

CHIEF SANDERS: A couple of things. First, as a former detective that got down and crawled around in the crime scenes, I would say that a lot of us that have specialized and that has changed because of that letter of law criminalists and those kinds of things that operate now.

The other thing is that I would like to suggest to you that as I listened to all of you identify your agencies and things, don't forget that 85 percent of all police agencies in this United States is like less than ten police officers or less than twenty. So oftentimes they are going to be the ones there that you cannot underestimate the importance of educating police officers or training.

I mean, I take exception to you. I don't think we sit around in our nice suits because I was a detective in South Carolina, and we wouldn't afford nice suits.

(Laughter.)

CHIEF SANDERS: But the ideas is that the points are valid and there was something else that I wanted to say that had to do with getting police - getting police involved in the program so that they know what's available and how to do it and those kinds of things. The thing that you got to realize as well is that the same as when you build new jails, we're very proficient at filling them up.

If, in fact, you don't work together so that we know what issues or what items, let's say. You got to make sure that when you tell us what you want us to do that we know that we got to work with the crime labs because, if not, we'll fill you up with stuff. I mean, we have police officers that will go to a crime scene and they'll take the whole damn house and send it to you.

Now, that kind of thing you got to be mindful of that so that when we try to establish a protocol it's got to be in such a way that we know what it is that we're supposed to be collecting. But I do want you know that we used to crawl around on those crime scenes.

(Laughter.)

MR. MORGAN: George, did I see that you had a question or?

MR. CLARKE: Oh, yes. I was just going to underscore that point that Dean made about different types of DNA analysts. I think it's important to have a broad spectrum. There's going to be those dedicated to working on rape kits frankly. And they don't have to and necessarily need to have that training at crime scenes and so on.

Then on the opposite side of the coin, if there's a difficult crime scene, a homicide that may or may not involve sexual assault, then it's going to be extremely helpful to have someone who understands, for example, crime scene reconstruction listing pattern interpretation and so on.

And those individuals, for example, and the most recent kidnap and murder of a seven-year old that we just solved out in San Diego, that was a highly experienced criminalist who became a DNA analyst and who now is able to put together all of those worlds.

So it's not as simple as though there's one perfect type in DNA analysts in, my view in forensics. It really involves a combination of a lot of different people.

MR. MORGAN: Barry?

MR. SCHECK: It's me or? I have echo what these last few speakers have said. But I think when you look at what all you've talking about, there is one class of people that are growing in the labs and those are people that literally are extracting, although some of them are going to be robots soon, in interpreting the DNA and applying the computer software and looking at the peaks, and all that should be done efficiently. But this other level of expertise that George correctly identified a bottleneck, this evidence screening, and going to the crime scene.

You know, from our perspective, looking a second time through at either old cases that are unsolved that I was doing with the city, which I guess, Ed Norris now who is the police chief in Baltimore, are old or post-conviction cases. I can't begin to tell you about the number of stains that are missed, all right, on Clothing.

The failure of people to take proper substrates to understand confounding results, picking out the right sample so you don't take the house and everything else, because preservation of evidence is going be an increasing power over time. What are you going to save? What are you not going to save? How much is it going to cost to do that? What is the legal requirements for that?

So there has to be a full-fledged effort and NIJ is the right agency to assist local people to create a class of people. I mean, frankly, they should be in part detectives. They can also be civilian-trained criminalists maybe through these, you know, college-base programs where you create a new class of people that understand what the technology is and how to collect and preserve the evidence correctly. It's a whole speciality that you're going to have to have to make this technology work correctly.

MR. MORGAN: Carl Selavka?

MR. SELAVKA: One comment and then a question. The comment would be that the new class of people is actually the old class of people. This DNA thing is a new class of people. So in the old days criminalists were broadly trained, were good at crime scene evidence, both at the scenes and in the laboratories. It's really about re-amplifying something that's been going on for forty years, and we just have to get back to that a little bit.

But I was intrigued by something that came up in George's talk about having two expert systems for convicted offender stuff. We heard about one. Is there another that would review data easily in the concordance of which would allow us to pre-screen 95 percent of those samples.

MR. HERRIN: I don't know. But the way we have TrueAllele setup is that it's designed for the computer to do most everything and to curate - think of an assembly line. So you go along in the factory. It does all the ladders. It does whatever it does. And then a person checks and making sure is this reasonable.

Then when you get to the end, a person reviews typically once it's been optimized, 15 percent of the data that's problematic. So it's really a person assembly line partnership. The computer isn't doing it all by itself. I don't believe in that. I think you need sanity checking.

If you wanted to, there are many modes in which you could run the system, including for almost no time having people look through the 85 percent where there are no problems. If you had to satisfy some rules, for example, that were all guidelines that were set. Because once you know the computer has run 25 checks on it, according to your SOPS, you're not going to find anything wrong with it. So if you need to look at it, that's okay.

You can also set it up so that there's a second review of what a person has done. The British are using it right now for most of their production work for the data banks. And they have - they run it through TrueAllele and then they have a double-human review of wherever the issues are.

So we really think of it as a configurable process, not as one piece of software, and you can set it up for almost anything you want it to me.

MR. MORGAN: On behalf of the panel, I'd like to thank everybody for their comments. And on behalf of the AGID-LAB group, I'd like to thank this distinguish panel for all of their work and presentations today.

(Applause.)

MS. HART: I'm afraid we're running a little behind schedule and I certainly contributed to that with all of my questions. So if we could possibly take a ten-minute break and try and move on to the next panel, if I can get you out of here on time.

(Recess.)