Identical Twins, Different Tissues, Disparate DNA

Tuesday’s New York Times reported that “new advances in DNA sequencing” are making it possible to “to tell identical twins apart.” 1/ The scientific principles that permit this feat are, by now, uncontroversial. Although cells are extraordinarily good at copying DNA when they divide, occasional mutations can occur. After all, human cell division entails replicating an ordered set of over three billion nucleotide base pairs. It is hard to be perfect billions of times in a row. When a mutation occurs in utero and becomes incorporated into the tissue line for one embryo but not the other, genomic sequencing offers a feasible way to detect the tiny differences between the children. 2/ So laboratories ought to be able to tell identical twins apart by their genomes. This genomic comparison may not work every time, but it should succeed in those cases in which the mutations were perpetuated in one twin but not the other. The article thus suggests that the technology of Next Generation Sequencing (NGS) could make criminal cases previously frustrated by the inability to assign crime-scene DNA to one and only twin “come back to life.” It opines that verified data on even a single base-pair difference would be "definitive."

However, the article completely overlooks an important complication. Merely discovering a few differences in the DNA of the twins themselves may be insufficient to exclude one twin and not the other as a possible source of the trace evidence. At the fine scale involved — a few base pairs out of billions — even a single individual might have sequence differences in different tissues. For example, an early mutation might have affected the cells that produce semen but not those that produce blood cells. In those situations, both twins seemingly could be excluded when comparing semen recovered in a rape case to known samples of blood or saliva from the twins.

This occurred in Commonwealth v. McNair, 3/ a multiple rape case in Boston. 4/ Contrary to what the Times suggested, semen from a condom used by the rapist did not simply match the base pairs in one twin but not the other. There were nine confirmed single base-pair differences in the saliva samples from the twins. The defendant's saliva DNA sample matched the sperm sample at only seven of the nine variants; there was a nonmatching pair of bases at each of the other two. His brother matched at two locations; he did not match at the other seven.

To make sense of this more complicated pattern of cross-tissue matches and mismatches for both twins, Eurofins called on a biostatistician, Michael Krawczak, then at Christian-Albrechts University, in Kiel, Germany. He performed two "Biostatistical Evaluations" to quantify the probability of mutations that could have caused the differences. The Times reported that “[b]ased on a statistical analysis by Dr. Krawczak, Mr. [David] Deakin [the prosecutor] told the court that it was two billion times more likely that the rapist's DNA belonged to Dwayne McNair than to his brother.” 5/ Under this interpretation of the data, the two mismatches for defendant were the result of mutations in the defendant's sperm as contrasted to his saliva -- a condition known as mosaicism. 6/

What Dr. Krawczak computed was a likelihood ratio 7/ — or rather a series of them. The computations rested on various assumptions about the formation of tissues in the early stages of human embryonic development and the application of an estimated, average mutation rate across the entire human genome to each base-pair site involved in the case. Although Dr. Krawczak testified on direct examination at a pretrial hearing that “the mathematics ... in the calculations ... is ... very straightforward, and it's not difficult,” on cross-examination, he conceded that his mathematics had never been presented in any other forum:

Q. So your likelihood ratio, depending on what you put into it, went from two billion to one to twelve thousand to one, yes?
A. Yes.
Q. And there was a third one that you put in some other numbers and that come out to something like seven trillion to one?
A. That's correct.
Q. So the ultimate likelihood ratio is wildly fluctuating depending on the numbers that you put in, yes?
A. That is correct.
Q. And a lot of the numbers that you put in are left to your discretion, yes?
A. They are based on literature information about mutation rates, and they are based on the information that I got from Eurofins.
Q. But there is no established formula among forensic DNA statisticians as to the methodology to use to calculate the likelihood ratio of one twin being more likely than another to be the source of a trace evidence sample based on the saliva genotype of the two twins, correct?
A. That is correct.
Q. Can we agree that the methodology that you created in this case, and that you had to create because there was no other methodology, is not generally accepted in the forensic DNA statistics community?
A. That is correct because it has not been put to the test of peer review, and that's correct, yeah.

The trial court’s response to the mathematics, and indeed, to all the scientific testimony, is confusing. On the one hand, Judge Linda E. Giles found that "the Eurofins testing and Dr. Krawczak's statistical analysis are based on generally accepted, valid scientific or statistical principles, including, inter alia, Sanger sequencing, NGS/MPS sequencing, and the likelihood ratio method." 8/ All these things, she wrote, "were unshaken by defense testimony." On the other hand, she found that the prosecution "failed to meet its burden with regard to ... four [out of five Daubert] factors." The sole factor the court deemed to be satisfied was the existence of "a known or potential error rate." But this "error rate" was nothing more than Dr. Krawczak's likelihood ratios.

Whatever its reasoning was, the court excluded the work of Eurofins and its statistical consultant from evidence. The state's interlocutory appeal of this pretrial ruling was rejected, and the case went to trial without the sequencing data. On January 12, 2018, the jury convicted Dwayne McNair of eight counts of aggravated rape and two counts of armed robbery.

But the conviction and sixteen-year prison term is not the end of the story. Presumably, what prompted science reporter Carl Zimmer to write about the case and other Eurofins' research into solving seemingly intractable criminal and paternity cases with identical twins is that "Dr. Rolf, Dr. Krawczak and their colleagues decided to write up a mathematically detailed account of their methods." The paper appears in PLOS Genetics. 9/ It presents "a novel mathematical framework to quantify the evidential power of the genetic data in cases attempting to identify MZ twin donors, based upon comprehensive DNA sequencing."

Such mathematical firepower is needed to cope with comparisons of a few base pair variants across different types of tissue. This situation brings to mind the aphorism that "If your experiment needs statistics, do a better experiment." The better experiment would be a comparison of the genotypes of twins from the same tissues. Of course, the reason that Eurofins had to work with samples of different tissues was not scientific. The prosecution did not seek a court order for semen from the McNair twins. It relied on the less intrusive alternative of saliva. It would have been more difficult to justify semen collection from an objecting twin. 10/

NOTE

1. Carl Zimmer, A New DNA Test Can Identify the Right Twin, N.Y. Times, Mar. 5, 2019, at D6.
2. Of course, one needs be sure that the differences are not mistakes in the sequencing test, but subsequent, highly accurate sequencing that zeroes on the few base pairs that seem to be different can overcome that technical problem.
3. No. 8414CR10768 (Mass. Super. Ct. 2018).
4. I consulted in the case for defense counsel, Robert Tobin, Jr., and received compensation from the Committee for Public Counsel Services for most of the time involved. At counsel's request, I submitted an affidavit and testified at the pretrial hearing as to the legal criteria for general acceptance and scientific validity and the application of these criteria to Eurofins' methods.
5. The Suffolk County District Attorney later issued a press release repeating that "Dwayne McNair was two billion times more likely to be the source of the DNA evidence than his brother." Suffolk County District Attorney's Office, Identical Twin Gets Prison in 2004 Rapes, Feb. 26, 2018.
6. W.D. Foulkes & F.X. Real, Many Mosaic Mutations, 20 Current Oncology 85, 85 (2013), doi: 10.3747/co.20.1449 ("it may be that all humans are mosaics—but that some of us are more mosaic than others").
7. By assuming the prior probability that the defendant was the source of the sperm was one-half, Dr. Krawczak converted the likelihood ratio into the posterior probability reported by the prosecutor. His testimony did not make this assumption clear. See David H. Kaye, How Experts (Mis)Represent Likelihood Ratios for DNA Evidence, Forensic Sci., Stat. & L., Sept. 24, 2017.
8. Memorandum of Decision and Order on Defendant's Motion in Limine to Exclude All Results of “Ultra-deep Next Generation Sequencing” for Distinguishing DNA from Identical Twins with Samples from Different Cell Types, Commonwealth v. McNair, No. 8414CR10768 (Mass. Super. Ct. Apr. 11, 2017).
9. Michael Krawczak, Bruce Budowle, Jacqueline Weber-Lehmann & Burkhard Rolf, Distinguishing Genetically Between the Germlines of Male Monozygotic Twins, 14 PLOS Genetics e1007756 (2018), https://doi.org/10.1371/journal.pgen.1007756.
10. In Commonwealth v. Kostka, 31 N.E.3d 1116 (Mass. 2015), the Massachusetts Supreme Judicial Court confronted an order for a DNA sample from a twin not suspected of a crime. Applying a "stringent" balancing test of individual interests versus the need for the evidence, the court held that an order for a twin's buccal sample was unreasonable under the Fourth Amendment. The prosecution's need for a DNA sample from the twin there was weak because it was not clear that there would be any "twin defense" and the state had fingerprints from the victim's living room and bedroom that matched the defendant's (and presumably could be used to exclude the other twin if necessary). In McNair, the prosecution had a codefendant who had pled guilty testify to Dwayne's participation.

MORE POSTINGS ON COMMONWEALTH v. McNAIR

• The Gordian Knot in Commonwealth v. McNair, Forensic Sci., Stat. & L., Mar. 11, 2019,