DR. CROW: I very much appreciate what the latest speaker said. It's very easy to oversell this magic stuff and assume that that will solve all of our problems. It is remarkable to have this kind of depth of knowledge, but I guess I should emphasize just as he did that just knowing the sequence of DNA doesn't tell us very much. It tells a lot, but it doesn't tell us all we really want to know about a person.

DNA is in the news now. Almost every day I pick it up and there is something about it. There is the announcement that the sequencing is now complete. A more correct announcement would be that it's approximately 90% complete, but that it's enough to put to use and it's already yielding interesting information and potentially important information.

It wasn't very many years ago that I had lunch or dinner with Robert Sinzere in Santa Cruz, and he was proposing sequencing the human genome, and I can remember my own thoughts that that's terrible, pie in the sky. It's incredibly expensive. It probably couldn't even be done, but I did think at the time that one or the other of two things would happen. Either it would be so slow and so expensive and so distracting that it would just fall under its own weight, which is what I really thought would happen, or that the techniques would change so fast it became a feasible thing to do. It's not as expensive, and, of course, the latter is what has turned out to be the case.

Right now the sequencing machines, which are really magnificent, can turn out DNA sequence at an enormous rate without any human intervention to speak of. So now the problem is putting these pieces together. It's almost as if you had a 10 foot long rope cut into approximately 1 foot lengths and then you have to put it back together again. This is done by matching the ends, but if you have a million pieces of DNA, that's roughly a million squared of possible matches, and it's just an enormous task to go through all of this. A computer can do it faster than people can, but it's still a major task, and my congratulations certainly go out to the mathematicians who have worked out algorithms for putting these very complicated things together.

It may take some time, and one of the reasons is that a lot of DNA that we have is really highly repetitive, so we have sequences here and there and elsewhere that are very much like others, and putting those together and into the right sequence may mean that finally getting 100% of this thing may take quite a while.

The task of our committee, however, isn't to solve the DNA problem. It's to predict what will happen in the use of DNA for forensics in the next ten years, and we've taken two year and five year and ten year intervals to look at. A few weeks ago I was invited to give a talk and predict what genetics would do in the next century, and in preparation for this I decided to see what people at the turn of this century, that is, the 20th century, what predictions they made.

One of the greatest scientists of the 19th century is Alfred Russell Wallace. He made two predictions -- he made several predictions. I picked the worst ones. He predicted that vaccination would be shown to be the fraud that it really is, and he quoted 50 pages to show how smallpox vaccination was a terrible scourge. The other thing he predicted was that phrenology would finally come into its own and would be the basis for behavior understanding. I offer that as a suggestion to our psychiatrist. If people as great as Alfred Russell Wallace can be off in their predictions, he gives one little pause before predicting too much.

There are a number of things that we're certain to map and other things that we think are quite probable. At a most recent meeting of this group our report was very severely criticized by Jennifer Smith, who said we should really throw out the population and statistical part of the report. I'm happy to say that I saw her about three weeks later. She gave a talk in Madison, Wisconsin, and I had the pleasure of introducing her. I might say I resisted the temptation to really take advantage of this situation and gave her a nice introduction, but it gave us a chance to talk about this afterward, and I found out that virtually all of her objections are substantive all right, but they're easily taken care of, and I think we have.

To mention a couple of them, the FBI as represented by her and others, too, were very sensitive to the possibility that if we talked too much about the future possibilities that are certainly going to be better than the present possibilities, that that somehow undermines the present.

So we made quite a point in the report to say that the present STR system and the 13 are really very good. That's not to say we won't get better things, but the 13 are institutionalized now, and changing them would be hard for administrative reasons and expense reasons even if something better comes along.

So our report isn't interpreted and we've tried very hard to make it such that it's not interpreted as in any way denigrating the existing procedures. We also made a rather snide remark, which I took out, about the British system. We found a 6 loci match and it was followed up and thrown away, but not by the forensics people I would say, but mostly by the news media. We've taken out any comparison with fingerprinting and we don't use the word "fingerprinting." There is a real question about whether fingerprinting is more or less reliable than DNA. But anyhow these are different subjects, and we've stayed away from it.

I wanted to call your attention to a sentence on Page 7, Paragraph 3, which I'll read in case you don't have it in front of you. Although this comes actually at the instigation of the FBI, I'm delighted to say the FBI did not dictate our report, but we're very glad for the input from that group.

"Although this report looks to the future, we emphasize that current state-of-the-art DNA typing is such that the technology and statistical methods are accurate and reproducible. Nothing in our predictions should be interpreted as casting doubt on the reliability and validity of DNA typing as currently practiced. Our predictions are consistent with the viewpoint that science is always evolving and will seek future improvements and alternative methods that are even better."

I think that takes care of the arguments that were raised at the last meeting, and I think we're ready to go ahead and keep this section of the report intact. There is a big investment of time, money, and energy in these, and they're likely to stay for a considerable length of time, probably the full ten years that's in our province to look at. It will undoubtedly be supplemented, but we don't expect it to be replaced within that period.

This report, as I've told you before, is not very much changed from what you saw at the last meeting. It's in two parts. The first 37 pages are intended to be user friendly. I've heard different opinions as to how user friendly it is, but I think most of it is that way. At least that's the intention. Then the rest, from Pages 38 to 71, is the more technical part for those who want to read the more technical parts, and then we include references, acronyms, and a glossary. I welcome your detailed comments at this stage, but if you have them, please do it soon. The working group is essentially finished.

What you see before you is the product of the editors from the Department of Justice, and we expect to work with them over the next few days and weeks to get a final version of this. I'm sure that there are many verbal things that can be improved, but I don't expect the content to change appreciably.

The main division of opinion within the committee -- I mentioned this last time -- is among statisticians. Statisticians have been notoriously argumentative for the whole century, and they haven't let up. Many statisticians try for exactness, try for a precise model, and they have preferred the likelihood ratio method of approaching analysis of the situation.

The likelihood ratio has very undesirable mathematical problems, but it's also less transparent that people who are going to use this in a more time honored traditional forensic community procedure, the probability of a match is still in vogue, of course, and I think will remain so, and our report doesn't deter that in the least. But there is a legitimate difference of opinion between those who want to be as precise as possible and those who think rough and ready calculations is perfectly good if you can show it doesn't make that much difference, and this is usually the case.

Before the last session we got a last minute report from the DNA advisory board, DAB, pretty much endorsing the conclusions of the 1996 NRC report as far as dealing with database searching and in other places where the different approaches led to different procedures. I say once again the different procedures perhaps never lead to different conclusions, but they're different ways of approaching them.

Our report actually presents both views. I might call your attention -- I was a little surprised myself to discover that I can use the two different procedures on the same set of data and ask how much difference it makes, and I was somewhat surprised -- it's on Page 67 -- if you look at these numbers at the bottom of the page, if you have it in front of you, what I did was take a 13 loci STR situation simply by assuming that all are equally frequent and putting in an appropriate number of homozygotes as would be predicted. If we have 5 loci at the very bottom line, you'll see that the probability of a match, everyone probably within the population is about 1.4 times 10 to the minus 15, a pretty small number, and if you use the best procedure, condition probability corrected procedure, it's 3.26 times 10 to the minus 15. That's a factor of two difference.

I say who cares? So people from the DAB and elsewhere who suggest we use simple approximations, it doesn't in any case make much difference. If you're dealing with numbers in the order of 10 to the minus 15, who cares about an error of 10 or even 100?

Well, I think I've said enough. The question of individualization, the FBI has a procedure -- of course, there is no way of ascertaining absolute certainty in a problemistic world, and no scientist and no statistician will ever say that something is absolutely certain. I think it would be hard to get a statistician to say that it's absolutely certain that all the air molecules in this room will fly out the door here and we'll suffocate. That's within the realm of probability, but hardly any of us regard it as probable enough to take seriously.
I think we are getting reasonably close with DNA to reaching probabilities that are close enough for practical purposes. One can say it's dispositive. The statement in the report says that individualization can be attained with a high degree of confidence, meaning not that it has a probability of zero, one or two extremes, but it has a high degree of confidence. That's what the DBA would like to say about the report, and it's our committee's viewpoint to go along with that.

The question of a database search, I'll say a little more than I said last time. I want clearly to distinguish it between a database search in which you're searching a database made up of convicted felons and a database made up more or less of a random sampling of the whole population. Among the convicted felons, the recidivism rate is high, and it's high enough to be probable.

So I see no real objection to just using the calculations directly based on the analysis from the database if it comes from the convicted felons. I'm aware -- and you people are in a much better position than I to discuss this -- I had made use of prior information when I made that statement, and yet the courts have been unwilling to take prior probabilities in advance. Whether you do this in your mind and then reach conclusions based on that or not, I believe that is a question for that group to discuss.

Sometime in the future if we reach the stage where our database consists not of convicted felons, but essentially of a random sampling of the population and a very large random sampling, then the prior probability of being guilty of this particular crime becomes very small, and therefore the probability of falsely detaining an innocent person becomes appreciable. A 1996 committee report had a procedure to take care of this, which is to multiply the probability by the size of the database. That usually doesn't change the condition I don't think. If the probability of a match is 10 to the minus 15 power and you have a database of 100 million, still the corrected probability of a match is still 1 in 10 million, again very strong evidence.

So I don't regard that even looking far to the future as a major difficulty, and the DAB has endorsed this particular procedure, but I want to emphasize, as I tried to do earlier, that there is a real distinction because of recidivism, because of a prior probability of there being a match or being the same person if it's gotten out of a database search of convicted felons than if it's gotten out of what may happen in the future, a much larger database. We're not in a position of predicting whether that will ever happen or not, but it might.

A few specific predictions. One I've already said, but I'll repeat is that the 13 are good and they're here to stay for a while, and the second one is -- sorry Barry is not here because this is almost for his benefit -- that the mitochondrial DNA will undoubtedly play an increasing role in the future, much more than it has in the past. It has some very valuable properties. One is that it's exceedingly sensitive in the sense that a very small amount of it can be revealing, but it's going to require some additional steps because whereby you can use a product to get different combination of genes, you can't do that with mitochondrial DNA, so that limits the power just depending on the size of the database, and I think there is every incentive to get a larger and larger database. Nobody can say just how by getting larger and larger databases can make this comparable.

At least one of the nice things about mitochondrial DNA I might say, mitochondrial are inherited from the mother and the Y chromosomes are inherited from the father. Why doesn't mitochondrial DNA and the Y chromosome DNA give you about the same kinds of information about an individual or about groups, and it turns out they're quite different.

Y chromosome inherited shows substantial differences between different parts of the human population within major races. The mitochondrial DNA shows much less. It isn't obvious why that's true, but anthropologists have told us why it's true, and that is in our human past. If we go back a few hundred thousands of years, this is an explanation of a matter of migration, and it's true in the human history that most of the migrations have been done by women.

It sounds like the opposite of what you would think. You would think the men are the explorers and go out, but what they go out for for the most part is to find women and then they bring them back. So the migration is by the women rather than the men. Whether this particular scenario is correct or not, the genetic consequences of such a thing in fact are. So that means mitochondrial DNA is particularly useful for identifying individual differences. We're in the business of picking out group differences. Otherwise we would make much more use of the Y chromosome.

What about twins? Right now DNA by itself can't distinguish identical twins. I don't doubt within the next five or ten years, though, it would be quite possible to identify identical twins. For one thing we have these wonderful things called expression chips that measure not just the genes that are there, but how accurate the genes are, and they've already shown that the same strain of mice, identical mice of different ages show different expression patterns; that two identical twins reared in different homes or even in the same home are certainly going to have different experiences, different infections. They will have all sorts of things that might change the way different genes are expressed.

So I'm sure this will be a way that will work. Another thing is to just pick genes like those that have to do with antibody formation, and they're likely to differ in pairs of twins and probably they differ in their infections by viruses, so maybe virus insertion parts could be used as ways of distinctions. Each of us carries our own different strains of bacteria, and maybe that can be used for identification.

Anyhow, my guess is that even today with a very concerted effort, very expensive, of course, we could in fact distinguish identical twins, but it will probably be routine by the time the ten years of the report is out. I don't know that I'll live ten years now to be held responsible for this. It might be convenient to get out of the way.

One thing that is predictable, though, and I think that within ten years -- and this is a solid prediction -- certainly sequencing or something equivalent will be so efficient that each person can be easily distinguished to be unique and maybe every high school student will have a DNA sequencing kit. I sort of abhor that possibility, but it's technically a feasible thing to do. Among the scientific advances over the next few years one of -- there are really two classes of them. One is improvements in STRs, and they are certain to come and they're certain to be welcomed. They will keep our same 13 loci, but will make us more efficient at handling them. There can be automation, for example, with less touching of human hands, and there is speeding up of the process.

So I think within the next five or ten years the STRs we will be using will be taking care of much more rapidly and much more efficiently and I hope cheaper than at the present time.

On quite a different level there are kinds of things being discovered under what we call SNPS. They're not likely to displace STRs because they have too few locus for locus. They're more valuable for genetic research right now than they are for forensics, but sooner or later there are so many of them they will undoubtedly come into the picture and they will be used.

Looking a little further into the future, one of the wishes of our committee or at least several members of you are as far as possible to avoid group identification. We're all conscious of the political overtones, social overtones, legal overtones of genetic profiling or racial profiling, so maybe the less we can use of racial statements in connection with that identification of people perhaps the better.

I think it's possible in the future to make DNA identification without too much specification as to what race the person belongs to, what major group a person belongs to. One thing to do is we have several different, five or so, ethnic groups in the state. You can test all five of them without asking which one the person belongs to. Another step which could be done is to have a uniform database cover everybody stratified by regular makeup in the United States and then use a large evaluator corrective factor to take advantage of the fact that there is more substructure in whatever one of the population is treated this way than otherwise.

The third is what I call a sib. method, which is to set your screen such that you could detect a difference if they are brothers and sisters or sibs. There are two rationales for this. One is if you can distinguish brothers, you can distinguish anybody else except twins, of course, and the other, though, is a technicality, but that is the relationship between two brothers is determined entirely by the genetic makeup of their two parents.

This is pretty obvious, and then so therefore we're dealing only with mental rules and no population genetic considerations come into this. That's not strictly correct because they do have an added effect, but the major factor is this factor of one fourth brothers and sisters are alike in their genetic makeup in any particular locust one fourth of the time, and that doesn't depend on any assumptions about the makeup of the population. So people can take some satisfaction in using this particular written method, the one fourth multiplied to several powers is the major constituent.

What about using DNA to identify the racial origin of a group? You could do pretty well right now with the 13 loci, but there is very little reason to do it. It might help in narrowing the circle of suspects, and that could be done right now. It can be done better with a large number of markers. We're more interested, however, in asking about the detection of particular traits, and right now there are a few traits that could be identified, bald or color blind. I'll soon run out of these because there aren't very many known single traits, but right now there are a few that can be used.

Undoubtedly this wonderful DNA project is going to turn up more genes, and perhaps within ten years it will be possible to make a pretty good description of a person typically from a DNA sample.

One thing that's interesting about this, and that is the defense statement that I didn't do this; it was my brother, and I don't know the brother, the brother is not a friend of mine, but for the reasons that I just said a while ago that's this one fourth similarity that comes in with brothers that doesn't happen elsewhere, and that means if you sample 100 unrelated people and there is one pair of brothers in there, the brothers may make a bigger difference than all the others put together.

So one thing that our report emphasizes is that one should be on the look-out for the possibility that within the people that are identified there are brothers. Usually this could be found out in other ways, but with 13 loci pure DNA analysis can almost always distinguish brothers.

That does have one legal implication, though, and it's going to turn up that one is going to find from a database search the brother of the person that you really want. Then what you do -- and I know that the states differ -- in particular, Virginia has a different procedure than some other states. In New Zealand there was a study of 10,000 people, 10,000 records. There were ten matches found for 6 loci. Of those 10,000 all but two of them were brothers. Some of them were twins actually.

That's all I want to say. I can't stop, though, without thanking two people, Lisa and Robin. They've kept me sane if nothing else.

MS. ABRAMSON: Any comments or questions?

MR. THOMA: I find the report fascinating. As much as I have been involved in this for ten years, I can't tell you how much new information actually is in here that I haven't seriously considered before, but --

DR. CROW: Let me say thank you and then I will prepare for the rest of the question.

MR. THOMA: At Page 23 you're talking about variations from Hardy-Weinberg ratios. It might be worth noting that there are some population subgroups -- you mentioned the major population subgroups, but smaller subgroups that vary from Hardy-Weinberg and just touch with it with regard to the major subgroup. That isn't that much that we found in specific subgroups. There is a major deviation from Hardy-Weinberg, just a suggestion.

DR. CROW: I absolutely agree, and this is especially true for Native Americans who have a tribal structure and quite different from that of the larger populations. We have made allowances for that. There is a discussion of that in the report.

MR. THOMA: At Pages 28 to 32, which is a difficult section for you to do because you're looking into the future obviously and you can't crystalize it completely, but I would be interested to know when and how you think there will be improvements in STR analysis to eliminate problems such as stutter and things like that. How soon and what is the most likely scenario for that? Obviously that's not an easy thing to do, but I was curious about that.

DR. CROW: We don't see anything very specific for obvious reasons, but I think I can say that I don't think there are going to be any major sudden break-throughs that suddenly eliminate stutter. This is a matter of attention to detail, doing everything a little bit better, refining all your methods, and it will grow incrementally.

One thing that is true it turns out that if you work with sets of five, repeats of five, they show as stutter, and if we were starting over again, I think we would make STR pentanucleotides rather than the tetranucleotides. That's one more thing that the system is set and I wouldn't want to change it, but if we were to start over, we'll do it.

MR. THOMA: To personalize the comments I'll send you a note on that.

DR. CROW: I urge all of you to do that, and to do it promptly because we would like to get this finished, but please make any suggestions about -- don't tell us to overthrow the whole report, but suggestions for greater clarity, wrong emphasis --

MR. THOMA: Thank you on behalf of the rest of the commissioners for this report. I really mean it's a lot of great information.

MS. ABRAMSON: Any other comments or questions or suggestions?

DR. FORMAN: These are the suggested covers. Everyone should have copies of the proposed covers, and I know that there was some discussion about the pamphlet covers. The group was fairly lively about that, so we wanted to get your input on the proposed covers that we have for the report. The report will be the same as the post-conviction report, and if you would like to discuss these covers, we would like to hear your opinions.

MS. ABRAMSON: First I would like a motion. Jim, why don't you make it instead of me that we approve and adopt this research and development and working group report subject to any editorial changes.

DR. CROW: I'll so move that.

MR. THOMA: I'll second that.

MR. ASPLEN: If I could as a matter of discussion on that matter just kind of the final question of whether or not anyone has anything other than editorial changes, anything that you would believe to be somewhat substantive, and the reason I ask is that pursuant to the Federal Advisory Committee Act we can publish something that are editorial changes from this point on; however, anything that would be of a substantive nature would have to be discussed in a public forum, and I just want to make sure that the changes that any of you are contemplating sending are just that, editorial changes.

MR. THOMA: No, absolutely not anything other than editorial, and I second the draft and I think that we've got a consensus on it.

MS. ABRAMSON: Is there further discussion on the report and we'll get to the cover later? Any further discussion? Then I will ask for a vote. Those in favor of so approving and adopting this working group report signify by saying aye.

(Chorus of ayes.)

MS. ABRAMSON: Those opposed? Then it's unanimous on the adoption of the report.

DR. CROW: Thank you.

MS. ABRAMSON: We say thank you to you.

So now we can get to the easier part maybe.

DR. FORMAN: There should be four choices. We can change anything.

DR. CROW: I don't want to be an art critic here, but I like this one and I like it for two reasons. This one that has the two little insets is all right, but that's more or less DNA of past, and I would like this to reflect DNA of the future. I don't know if anybody knows what this is. It's a chip, but it's not the kind of chip that is being worked at the cutting edge of this particular technology, and if we're going to have a picture of a chip, I suggest -- Lisa knows this -- that you get in touch with the Harvard-MIT group and get a picture of what theirs is. I think they are probably closer to something that would be practical for routine use.

MR. THOMA: My only comment was going to be for the same reason. This particular one I was not as interested in as the others, and basically I would agree with Dr. Crow that this of the other three would probably be the most --

MS. BASHINSKI: My only problem with it is the color is a little --

DR. FORMAN: Boring?

MS. BASHINSKI: Yes. I'm looking for the word.

MR. THOMA: Boring is fine.

MS. BASHINSKI: I don't know if there is another color combination.

DR. FORMAN: Are there colors here that you like better?

MR. THOMA: I like these colors, but I don't like anything about it.

DR. FORMAN: This picture and a better chip.

MS. ABRAMSON: I hear a consensus about this tan and blue, but with the purple and darker color and a different chip. I see a lot of nodding of the heads, those that are artistically inclined. Hearing no other --

MR. REINSTEIN: We really like our colors better on the post-conviction report. They were a little more exciting.

MS. ABRAMSON: Anything else to come up for the research and development working group report?

MR. ASPLEN: If I could just add my thanks also from a staff standpoint, the details of putting something like this together are tremendous, and we've said this before, but one of the many defining moments of this Commission's work in terms of its ultimate success -- and I do consider this to be a large success -- was, quite frankly, the lunch that Dr. Forman had with Dr. Crow where we asked him to participate in this meeting, and he agreed. As soon as that happened I should have seen it coming away from that meeting we were just overjoyed, but as the ones responsible for administering and trying to help this group be successful from the staff we want to thank you. You cannot ask to work for a kinder, easier going person yet as thorough as you can be than Dr. Crow, and we greatly appreciate that.

MS. ABRAMSON: With that we'll take a break from 11:00 to 11:15, and then we're going to hear from Dr. Paul Ferrara from the laboratory funding working group.

(Whereupon a brief break was taken.)

Previous	Contents	Next
Back to National Commission Main Page