DR. CROW: All right. Well, my privilege, stepping on to the -- in the agenda, is to introduce myself and to give a report on the technology group.

I would like to tell you what we've done, what we're planning to do, and at the end, or even in the middle, have suggestions of other things that we ought to have been doing and have not done or ought to be doing and will do later.

To start off, it's remarkable to me, and I trust to everybody else, how this field has developed.

It was about 14 years ago that Alec Jeffreys first discovered the principle on which we now operate, these repeat units that are so useful, and the FBI got into the act about 10 years ago, and in the years from 1985 to '95, the field was dominated by VNTRs, variable number of tandem repeats, just a very powerful system, as we all have learned.

There are many alleles per locus, and combining four or five or six of these can yield astronomically small or the inverse of astronomically small probabilities.

But there are also features about it that weren't the most desirable.

One was that the number of gradations of sizes of these are too small for the resolving power of most of the units that are -- most of the techniques that are used, and so, it's been necessary to group them into groups and treat the groups as units.

That inevitably introduces statistical complications, and it's worth saying, too, that he statistics that have been used routinely for this are certainly not optimum from a statistical standpoint, although they are much simpler than what would probably be optimum.

In any case, we're moving pretty rapidly away from this, and rather soon, these will be of historical interest rather than current interest.

One other problem is that, as first use, these depended upon radioactive probes, and that had two disadvantages.

One was that it took a very long time. Usually it took a week or so to analyze each locus, and then months could go by before the result was obtained.

It also meant one had to dispose of radioactive byproducts, which is always a nuisance.

The radioactive probes were replaced by chemical tests, to a large extent.

The other defect about this, though, that led to its largely being replaced is that it requires relatively large amounts of DNA, and then, about the middle of the last decade came this wonderful technique of polymerase chain reaction in which it's possible to do essentially what a cell does and multiple a small tiny bit of DNA up into almost any amount that you want, and that means that one can start with exceedingly small amounts of DNA, pico-gram amounts, nanogram amounts at least, and that the -- furthermore, the units that are being used now, the individual repeat units are considerably smaller.

The total number of them, or the total length, is considerably smaller.

That's required, but the fact that, if you want to use PCR amplification, the molecule can't be too large, but it also means that there are a large number of such loci, almost an unlimited number, and it's simply a matter of taking the time to find new ones and get them adapted to useful work.

Not only can this take up very small amounts of DNA, the DNA doesn't have to be particularly good. By "good," I really mean long and not broken up into pieces.

This means that DNA that's been treated badly for any or all reasons and is broken up into smaller pieces -- it's still useful for this technique or the technique is still useful for it, and there's no binning or bunching required, because each of the alleles is distinct.

There are exceptions, of course, but for all practical purposes, this is true.

So, now, with the -- we're really in the midst of the STR time-scale, and the question for our committee to consider is how long before something still better comes along, and I'll talk about that in just a moment.

Along with STRs, though, there are two others that I want to give particular emphasis to.

One is mitochondrial DNA. It's available for even smaller amounts, the reason being that, whereas with ordinary DNA and the chromosome, there's only one unit or two per cell, in mitochondria there are hundreds, thousands of these particulars in an individual cell, so a very tiny number of cells, in principle even one, might yield enough particles to be analyzed.

They have the advantage and the disadvantage of being transmitted exclusively through the female line. That gives you the advantage of tracing certain kinds of ancestry. It means certain other types are indistinguishable, like two children of the same mother.

Combined with the other systems, though, this can be extremely valuable and has already proven to be so.

Corresponding to mitochondrial DNA and marking the female line of descent, there is the Y chromosome, which follows the male line of descent. For a long time, this is pretty largely unavailable for use. The Y chromosome is mainly inert.

It turns out, though, it has a lot genetically uninteresting but forensically very interesting loci that can be used for this kind of marking purpose.

There are -- the CODIS system now has 13 loci in it, and I'll say just a little bit about the power of the new systems.

If we have just 12 loci, one less than the FBI asks for, the average match probability is about 1 in 700 billion.

It can be smaller or less, but this is the average on the whole, and actually, with the black population, about 1 in 6 trillion, on the average, would match for 12 loci. It's interesting, and perhaps useful at times, that the black population is more variable than either of the other three that we ordinarily have data for.

I want to think about the future, and it's not very far in the future, in which we can, to some extent, avoid one of the most troublesome problems, and that's the necessity to classify people by race in order to do the DNA testing.

The original DNA committee, 1992 committee, developed what they called a ceiling principle just to get around this problem. The principle is roundly criticized, and it didn't succeed in doing what it was supposed to do, but the motive was quite right.

It was to see if one could have a system in which it was no longer to keep -- no longer necessary to keep databases for blacks, whites, hispanics, and orientals, and other possible subgroups, too.

The number of loci is now large enough that the probability of two individuals that are very closely related is still small.

So, one way to look at this problem and a way that the committee's been doing is to say -- is to ask ourselves, can this system distinguish between brothers or sisters or brothers and sisters?

The reason for asking it that way is that the relationship between sibs is really of a different type than any of the other relationships that we ordinarily deal with. Two individuals that are sibs have a chance of one-fourth of being identical in their genetic makeup at any particular locus, and that one-fourth is unrelated to gene frequency, allele frequencies, or anything else. That just comes from the nature of the relationship of sibs.

That's not true of any other close relatives. It's true of double first cousins, but I'm not going to worry about them.

That means, then, we have this one figure of one-fourth which provides a minimum, and then added to that is the smaller quantity that's based on -- that does depend upon the frequencies of the genes in that particular group. I shouldn't say genes, the frequency of the types that we're dealing with in that particular group.

So, that means that, much of the time, the probability of a match between sibs is maybe 25 or 35 percent, or 35 or 40 percent, but the 25 percent is fixed and doesn't depend upon anything else. You don't have to know anything about population structure.

That means that, if we calculate the match for sibs, it's not going to be perfect, because this one-fourth is not the whole picture, but it's a large part of the picture, and that means that the comparison between a calculation made on the black population, the white population, and the hispanic population will be in much closer agreement for sibs than they are for unrelated individuals or other degrees of related individuals.

So, how likely is -- how many loci does it take until we can reasonable expect to distinguish sibs with the same kind of precision that, a few years ago, were used to separate individuals who were unrelated?

With 21 loci, which is not unrealistic nowadays, the probability of -- the average probability, now, of a sib matching in the black population is about 1 in 1.4 billion. For the white population, 1 in about 7/10ths of a billion, and roughly the same amounts for Asian and Hispanic.

So, we can achieve what I'd regard as a major step forward and simplification in this of having to pay less attention to racial subdivisions than has been necessary in the past, and I think that will avoid -- at least avoid some political complications, as well as social and ethical views that necessarily arise in this, but that doesn't mean that the grouping of data by racial and ethnic groups won't still be used, because one issue that's certainly going to come up, is up, has come up, and that is can we do the problem the other way around?

Instead of asking what the probability of a certain genotype is or profile is, given the genetic -- given the racial makeup of the individual, what about the -- looking at the problem the other way around?

Given the DNA profile, what can we say about the person that contributed to that profile?

Well, right now, one could write a likelihood ratio for any particular case and say that this profile is 10 times as likely to have come from a white individual as from a black individual, and perhaps that's useful information at the investigatory stage.

If that racial, instead of being 10 to 1, were 100 to 1, it would be extremely useful.

We certainly have nothing now that corresponds to an absolute identification, but with the -- in the course of the next five or 10 years, which is the period we're supposed to be looking forward to, the likelihood of making a pretty good guess as to the ethnicity of a person that left a DNA sample is here, and I assume that police departments are quite interested in having that particular piece of information.

A little bit about the new techniques.

There are certain -- there are certainly going to be things that are better than anything we have now. They're being developed by all kinds of research programs. They'll be a natural outgrowth of the human genome project, which keeps turning up new loci, and it also turns up new techniques.

We can expect radical changes, but whether -- how rapidly those ought to be instituted is something that I think that calls for some debate and some thought.

There are a number of things, though, that can happen that don't change the essential system but will simply improve it.

More automation is certainly one.

Miniaturization is going on and will continue to go on, and speed.

With miniaturization and speed, one can move toward a time in which one can do much more at the crime scene itself, rather than waiting for laboratory developments later.

Sometime within your lifetime, probably not mine, this might become a practicality.

Some of the techniques that our group has discussed and will write a description of -- the way we planned the report is to write a fairly general report and then with appendices for technical details, technical details regarding each of these particular methods and the statistical procedures, too.

Some of the methods are capillary electrophoresis, which is already here, beginning to be used more widely, various kinds of micro-technologies.

Much of the DNA now depends on sets of four in repeater units.

Pento-nucleotide repeats have the advantage of not having quite so much noise, stutter is what they call it, and maybe they'll be moving more in the direction of these compared with the ones that are now based mainly on four units.

They're SNPs, single-nucleotide probes, which go directly to the DNA sequence. That would, in many ways, be an advance, but it would be a fairly major change in the technology.

So far I've discussed things that depend almost entirely on chemical procedures.

There's also coming, I'm told -- I know very little about this -- mass spectroscopy, which can use physical measurements to get somewhat of the same information. Presumably, it could be done much more rapidly, and I'm sure it's here to -- will come in the future.

And then there are these various kinds of chips that simply miniaturize the process. How soon they will impinge on this system isn't immediately clear, although we're trying to unravel it.

There's sort of a policy question here or maybe economic question, and that is this, that the investment in CODIS is a considerable one, and we don't want to develop those 13 loci and then announce next year that they've been superseded by things that are better, and even if things are better, I don't think we would want to make that kind of a decision.

So, there needs to be a length of time that a standard procedure is used before it's replaced.

These are not necessarily mutually inconsistent possibilities.

The CODIS system could continue for the foreseeable future. More loci can be added or more different kinds of techniques but especially more loci, and then, I think I would foresee, but we'll see if it happens this way, that the CODIS database would be used more for identification than vital evidence on a particular case, for which there will be an abundance of other kinds of tests that can be applied to it.

Another kind of issue that we certainly need to -- and have said a bit about, will be saying something about -- is the -- is going to be DNA evidence from sources other than human.

There are already court cases in which plants have been used and domestic animals and, in some cases, wild animals. I read one about mooses.

These certainly don't easily conform to the kind of standards that we have set for the human population. Nobody has gathered the kind of large databases that we have.

Also, these populations, especially domestic animals, dogs and cats, are complicated by extremely high levels of inbreeding, which means everybody is somewhat related to everybody else.

So, it would really require a more careful kind of attention if we want to do it in the same -- with the same level of rigor that's been applied to the human population.

I think, however, that technology will help with this problem, too, essentially by supplying ways of analyzing enough DNA, enough different loci, that problems like inbreeding and a small database become less important, but how soon that will come is something to be addressed.

I've used my time, but I want to -- a purpose of this -- I told you roughly what we have in mind doing and, to a fair extent, have done. What else should we be doing that I didn't mention?

Yes, Barry.

MR. SCHECK: As I sit here, I can't remember exactly where NRC-2 came out on the issue of mixtures and whether or not, in particular, a statistic or a series of statistics or a series of conditional probabilities ought to be offered to the fact-finder about mixtures.

One issue that appears to be emerging is instances where the prosecution will do a DNA analysis using any of the techniques, particularly, though, I suppose, of greatest concern, STRs, and will then tell the fact-finder the -- this is a mixture of two or more individuals, are contributors to this mixture, and the defendant can't be excluded but offer no statistics whatsoever.

I can't remember how NRC-2 came down on that, and I was wondering if your committee had given any thought to the whole issue of addressing mixtures.

DR. CROW: Well, we're certainly going to address it, and I can tell you what NRC-2 came down on, having written that section. We essentially copied the procedures from Abbot, who really tells you how to make the calculations, and we provided sample calculations in quite a number of simple cases, but that's a long way from exhausting the possibilities.

MR. SCHECK: But the question I had specifically was -- I remember that.

DR. CROW: That's all I remember.

MR. SCHECK: But the issue was that, with respect to NRC-1 and NRC-2, I mean the position has been that you give numbers. It's not a question of not giving numbers at all.

DR. CROW: Yes.

MR. SCHECK: That's an essential point here, because what I'm pointing out is that there is an emerging trend simply not to give numbers, which I thought was an issue that had been resolved long ago but apparently isn't.

DR. CROW: We probably didn't say that explicitly but only by implication.

MR. SCHECK: Yes. I could show you a lot of cases that are proving this.

DR. CROW: This quickly gets complicated, I don't have to tell anybody, and it requires a very skillful probabilist to work these things out.

Some of that's been programmed now, and there are some machines that are cleverer than the average calculator, and I do foresee the purely statistical problems will become more formalized, more automated, and I suppose it's our opinion that we should be as quantitative about this as anything else.

MR. THOMA: Jim, one question about something that you did bring up, as opposed to something you didn't. What time shelf-life are you thinking of? I'm thinking of STRs,for example.

We're still working out some of the kinks with the noise or stutter that you're talking about. What time frame are we looking at as far as a particular technique that would be something we'd want?

DR. CROW: We've talked about it. You're not expecting a definitive answer, and I don't have one, but our time-scale, our instructions, really, have been to look at five years and 10 years.

Actually, we've extended this. We're going to try to see what one can expect in two years, five years, and 10 years, and essentially make guesses about that. So, that's our -- that's the time frame we're thinking about, and everybody in the room knows that we can only guess.

MR. CLARKE: Is the concept of these newer techniques to continue to use STR-type loci?

DR. CROW: Both. There are almost daily improvements in STR techniques, and more automation, newer loci are added that have properties that are better than the existing ones and so on, but some of the other things I mentioned are really different, and they would replace STRs.

MR. CLARKE: Because obviously, as you mentioned -- you mentioned the concern of what would happen with any national database and so on. So, would it become obsolete, or could it be done in conjunction, as you mentioned? It sounds like a logical direction.

DR. CROW: Not only the CODIS database has an interest, but many laboratories, of course, get tooled up to use the technique and don't want to change it capriciously.

So, I think our committee will favor some conservatism in this regard. After all, the STRs are pretty good.

MR. GAHN: When you talked about maybe SNPs coming up in the future, what sort of has concerned me over the years is that it seems like the scientists kind of run where all this is going and law enforcement then just gets behind that and makes their changes.

Speaking of my own preference, I'd rather try an STR case with gel electrophoresis than capillary electrophoresis, but it's the scientists who are going to run this.

What's going to happen when SNPs seem to be -- that's the choice of the scientific community? Won't that be a pretty big change for everyone in law enforcement?

DR. CROW: I think it would, and I was almost going to say the opposite. It seems like the scientific community has not been too successful in moving things as rapidly as it would like to, because there's considerable inertia in the legal system, and it's good.

I think we could just count on developing one new technique after another, and sooner or later, some will come along that are enough better that it will be convincing to the testing laboratories and to the legal community that that replacement ought to take place.

I don't know how to be anymore specific than that. Every week in the New York Times, there's a new article about new technique of some sort or a new miniaturization or a new robot, but I -- well, I'm expressing too many opinions on behalf of the group that hasn't expressed any yet.

MR. ASPLEN: Dr. Crow, the next meeting of the American Judicature Society, which is to be held about a month from now, is addressing -- I think the title of the conference is "Whose Genes Are They?" and one of the discussions that's going to be held is -- revolves around the idea of genetic defenses, if you will, not DNA used to identify perpetrators but, rather, the other side of the coin, and that's can we use DNA to determine genetic issues in an individual that would lead to a defense of some sort?

Has your working group thought about looking into any of those issues, issues aside from the law enforcement application for identification purposes?

DR. CROW: I think it's fair to say we haven't. We've talked almost entirely about the techniques themselves, and maybe we ought to ask about such questions as that.

MR. ASPLEN: I throw it out there as something that the Commission may want to think about.

DR. CROW: I think there's a jurisdictional question here. Our committee claims no expertise in ethics, for example. On the other hand, the Commission as a whole does, and some of these questions that -- maybe are better raised by the technology and answered elsewhere.

MR. SCHECK: Actually, I think that one way to get at that -- because I think, actually, Chris, you're raising a very important point -- is looking at the committee, Dr. Crow, as it's constituted. It might be useful, at your next meeting, or maybe in conjunction, even, with the legal working group, to find some medical geneticists, because I think that the issue is going to arise in the context of those geneticists that are looking at psychiatric conditions that are related to genetic antecedents, and you know, there's quite a bit of that going on, and I think, actually, it does have enormous importance, and not just in the area, incidentally, of raising psychiatric defense to criminal behavior, but I think we have to consider very seriously civil commitments based upon genetic profiling of mental problems.

In particular, I'd point out that the statute approved by the United States Supreme Court in the case of Hendricks v. Kansas, which permits sexual predators to be civilly committed indefinitely, has language in it that deals with, quote, "inherent disorders," unquote, which leaves a big opening for genetic proof, and there are -- there is some legislature -- legislation being proposed with respect to sexual offenders in many, many states.

DR. CROW: I'm sorry that Phil Reilly isn't here. He would have quite a bit to say about this. Is he coming?

MR. ASPLEN: He is coming. He had an engagement this morning but will be here.

MR. THOMA: Just to add to what Barry's saying, in California, as a follow up to Hendricks, we've got a couple of experts statewide that are testifying for the prosecution for the extension of these cases along those lines, and DNA has not been part of the playing field, and I think, actually, we should discuss this at an early juncture in the future, when Phil's available.

MS. BASHINSKI: I think, also, given our discussion that we're going to be having about collecting samples on a much broader basis, this really is an issue that will come up and will be very relevant to that discussion.

So, to the extent that these issues are being raised and they're really red herrings, that's one thing. We should identify that, and if there are really valid concerns that this information will soon become available and potentially misused, then we need to factor that into our discussion.

DR. CROW: I certainly don't want us to be blind to possible misuses. I don't want to be paralyzed by fear of misuses either. So, it's tough.

MS. BASHINSKI: Exactly.

JUDGE REINSTEIN: I talked last week to Dr. Mallott from the University of Maryland, David Mallott, and he's been doing a lot of research on pre-disposition. He predicts, in the next year-and-a-half, two years, in conjunction with the human genome project, that they will identify -- I think it's a gene -- it's predisposition toward impulsivity gene.

I know somebody else was doing some work on the novelty-seeking gene, you know, and all this has to do with predisposition of violence.

DR. CROW: It's time to move on to the next subject, but one thing, in my long career, has been seeing oscillations up and down, with an unpredictable period and enormous amplitude, between blaming everything on heredity at one time and nothing on heredity, and I hope we're going to come to some sort of reasonable balance.

I think it's very likely we're going to find genes that have some -- identifiable genes that have some role in impulsivity, but I certainly wouldn't want to call it an impulsivity gene at this stage. It probably acts -- I don't have to say this -- it acts in a complicated way, and I'm not sure that these predictions are -- it's going to be quite a while before they're a sharp as the newspapers would imply.

JUDGE REINSTEIN: What he said was he predicted that it would not be utilized as a defense in the next five years, but maybe 10 years out, that it would be used in mitigation, where the evidence standards are different, and you talked about the environmental factors that would trigger the gene.

You know, you may have it, but if you don't demonstrate alcoholism, maybe if you mix alcohol with that, it will trigger a chemical reaction to, you know, enhance the quality of that gene. I'm not sure.

DR. CROW: You mean that a person might use -- say that I can't help it because my genes dictate it, I had to behave this way, and use that as a defense.

JUDGE REINSTEIN: They call it the my-genes-made-me-do-it defense.

DR. CROW: Right. Well, you know, we went through this 10 or 15 years ago with the XYY syndrome.

Now, that's a very sharply defined trait, and there was a clear correlation between having an extra Y chromosome and ending up in a penal institution, but that says nothing at all about all the causal chains that led onto this.

Let me ask that we move to the next item on the agenda, which is yours.

Previous	Contents	Next
Back to National Commission Main Page