By Nicole Bronnimann, SLS JD expected ’18
In October 2016, a jury in the District of Arizona found Anthony Shirley guilty on six counts of aggravated sexual assault. Judgment at 1 (02/17/17), United States v. Shirley, No. 4:14-cr-00622, (D. Arizona Apr. 10, 2013). The jury came to their verdict without the aid of DNA evidence and yet an analysis of the Y-chromosome DNA left by the perpetrator of the crime matched Shirley. This evidence was not introduced against him—why?
The answer has to do with population substructure, the federal rules of evidence, and the role that DNA has come to play in the courtroom. In television courtroom dramas, DNA evidence is often portrayed as a “smoking gun,” conclusively determining guilt or innocence. Generally, this evidence refers to an analysis of short repeated nucleotide sequences of autosomal DNA, the so-called CODIS markers. However, some types of DNA evidence, while still probative, make guilt only mildly more likely. In this case, the government intended to introduce against Shirley an analysis of short repeated nucleotide sequences on the Y-chromosome. Instead of a “smoking gun,” this evidence is more analogous to a “matching shoeprint.”
Judges are often called upon to decide which evidence is fair to expose to the jury. In the context of DNA evidence, they must ask themselves two questions: 1.) Is the science underlying this type of DNA evidence reliable? and 2.) Will the jury have the nuanced understanding necessary to weigh it properly, considering their preconceptions about the significance of DNA? In other words, within the category of DNA evidence, will the jury be able to tell the difference between a smoking gun and a matching footprint?
In U.S. v. Shirley, at the pre-trial Daubert hearing, experts hired by the prosecution and defense offered differing testimony about the merit of the evidence in light of three principal issues. Considering these issues, the judge ultimately ruled the evidence inadmissible.
Issue 1: Analysis of the Y-Chromosome as Opposed to Autosomal DNA
Y-chromosomes are generally passed from father to son without change. The only changes occur through mutations, which happen, on average, only every several generations. Brothers share the same Y chromosome. Male cousins connected via their fathers share the same Y chromosome. If you are a male reading this, you can expect to share the same Y chromosome as someone in your community whom you have never met and would not call a relative. This is unlike the 23 autosomal chromosomes passed by either parent to a child, which are combined to form a unique profile of 23 pairs in every person (who is not an identical twin). The Y-chromosome’s wholesale transfer means it can be helpful for anthropology— in tracing how cultures have interacted—but it is less helpful for forensic purposes. While a negative Y-match is as powerful for exculpatory purposes as a negative CODIS match, a positive Y-match does not carry the same evidentiary weight as a positive CODIS match.
Indeed, if a forensic scientist were given the choice of analyzing the Y-chromosome or the autosomal DNA from a sample left by a perpetrator, she would never choose the former. The statistical probabilities of sharing a Y-profile are astronomically greater than of sharing a profile for the 13 CODIS markers used to analyze the autosomesThe odds of a positive CODIS match are often one in many billions or even trillions. With Y-chromosomes, because of the degree of sharing, smaller numbers can be expected— like 1 in 35.
Therefore, forensic scientists usually analyze Y-chromosomes only when they have no choice. For instance, in sexual assault cases with a male perpetrator and a female victim, there can be an overwhelming amount of female DNA in comparison to a relatively small amount of male DNA. Analysis of the Y-chromosome allows a forensic scientist to isolate the perpetrator’s DNA. An analysis of the short tandem repeats of nucleotide units (STRs) are then analyzed at different loci on the Y-chromosome to provide a profile. This profile can be compared to the defendant’s Y-sample. A mismatch excludes him. A match means he cannot be excluded but does not provide strong evidence that he was the perpetrator.
The match is then contextualized by comparing the perpetrator’s Y-profile to those contained within a database, to see how often a match would be expected. For instance, in this case, the government wished to admit as evidence that a match would be expected in 1 of 76 Native Americans. Government Memorandum Re: Motion to Re-Open Daubert Hearing and Preclude DNA Evidence at 13 (07/27/16).
Issue 2: Population Genetics and Available Databases
The defendant in this case was Native American, belonging to a tribe based near Tucson, Arizona. While forensic Y-STR evidence is now generally accepted, debate remains as to its use for Native American populations. As is typical, the database used in this case contains thousands of samples and divides them by race/ethnicity: Caucasian, African American, Hispanic, Asian, and Native American. The Y-sample from the crime is compared, at a minimum, to the Y-samples of the defendant’s racial/ethnic background to provide a matching probability within that population.
Racial contextualization for matching probability is also done for the autosomal evidence. The Evaluation of Forensic DNA Evidence, National Research Council (1996). Since the beginning of forensic DNA technology, scientists have considered how statistics are affected by the racial category to which a DNA sample is compared and how the racial categories within a given database are defined and validated. Population substructure—or the features of a population which result in significant variation of allele frequencies across individuals in that population—can affect match probabilities. When populations are isolated for many generations, and do not mingle with other populations, their gene pool may not contain the natural variation necessary for random probability analysis to be accurate. The genetic profiles of members of this isolated subpopulation may instead need to be compared only to other members of that subpopulation to avoid making a profile appear more rare than it is by comparing it to a population that diverged before the many generations of isolation.
However, because the probabilities of matching are so low for autosomal DNA, the influential National Research Council (NRC) report of 1996 concluded that population substructure did not meaningfully detract from the value of autosomal DNA evidence. Id. at 121-22. With matching odds like one in a trillion, the variable of population substructure is not going to significantly alter the weight of the evidence. Therefore, in cases of inadequate databases, the NRC report on autosomal DNA evidence suggested providing the matching probabilities of multiple races and introducing whichever was most favorable to the defendant. Id. However, that conclusion may not extend to Y-chromosome STR analysis, which has much higher matching probabilities. With higher odds, the effects of population substructure could have a more significant effect on the weight of evidence, if for example the variable could make the difference between 1 in 35 and 1 in 76 odds.
The point of disagreement between the experts in this case was the relevance of population substructure to the statistical match probabilities on the Y-chromosome of a Native American defendant.
There were slightly over 4,000 Native American Y-samples in the US-YSTR database from which the government derived its 1 in 76 match probability. However, the samples were not broken down into specific tribes. Thus, a Y-sample from a member of the Lakota tribe in North Dakota and a Y-sample from a member of the Tohono O’odham tribe in Arizona were both classified as “Native American” in the database. The defense and prosecution experts disagreed about the effect of this pooling on the statistics.
The defense expert argued that the amount of population substructure among tribes was significant to the point that that the category of “Native American” in the Y-STR database was too broad to provide a precise statistical basis. The database contained few if any members of the defendant’s tribe. By comparing the defendant’s Y-profile to an inadequate or misrepresentative sample, the profile could seem more rare than it really is. The result, in the defense’s view was to “manufacture diversity” and in this way, “to frame the suspect.” Transcript of Motion Hearing (07/16/16) at 20. The defense cited a paper by University of Arizona researchers Michael Hammer and Alan Redd, indicating greater population substructure among Native Americans than within other racial/ethnic groups. These researchers recommended the expansion of tribe-specific Y-STR databases for Native Americans.
The prosecution expert argued that racial/ethnic categorization in Y-STR analysis is not highly relevant. The defendant in this case, the expert noted, was a case in point because while belonging to a Native American tribe, his Y-chromosome was of European origin—indicating that his ancestors had intermixed with other populations that have come to the area. Therefore, in the expert’s view, subdivision within the Native American subgroups was “a totally moot question in the context of this specific haplotype.” The defendant’s Transcript of Motion Hearing (07/16/16) at 134.
The expert further concluded that because racial phenotype is not a reliable predictor of paternal ancestry, any breakdown of Y-STR evidence by race was unnecessary. However, the database contextualized the match by breaking it down into probabilities by racial category. After the expert’s testimony, the government sought to offer into evidence the odds of a match in Native Americans (1 in 76) as well as Caucasians (1 in 63) and Hispanics (1 in 76). Order (09/30/16) at 1. This reflected the standard recommendation that the NRC report makes for cases of inadequate databases in the case of autosomal DNA samples. However, for the Y-STR evidence, the court rejected including any matching probability because of the “limited database.”
Problem 3: Relevance vs. Undue Prejudice
Regardless of the nuances of population genetics, a match is a match. At its most basic level, the evidence showed that Shirley could not be excluded as a contributor of the perpetrator Y-sample. Wasn’t that still relevant for the jury to hear? The federal rules of evidence do not define relevant evidence as that which makes a material fact more probable than not. The rules determine evidence to be relevant if it makes a fact more probable than without the evidence FRE 401. Y-STR, in other contexts, has been compared to footprint evidence. We know that a footprint alone does not prove guilt (multiple people may own identical pairs of shoes) but a matching shoe is still relevant to the probability of the defendant’s guilt.
However, Rule 403 of the Federal Rules of Evidence limits the admission of relevant evidence by giving the court discretion to exclude evidence whose probative value is outweighed by the “unfair prejudice” that it would cause. The judge in this case cited the inadequacy of the database and the possibility of exaggerated statistics as justification for finding “unfair prejudice.” Moreover, he did not wish for the trial to devolve into a “battle of the experts” which would “leave a jury with little guidance in how to evaluate the weight to be given to the DNA test results in this case.”
While not mentioned in his order, it is conceivable that the judge also weighed the so-called “halo effect” of DNA. The mere acronym “DNA” may strike into the minds of jurors a certain confidence and certainty that makes them unable to objectively limit Y-STR’s probative value, especially in a case with relatively low probabilities. From TV crime shows to exoneration projects, DNA is seen as be all and end all of evidence.
On the flip side, this also means that jurors may expect to see DNA evidence in a sexual assault case and therefore make a negative inference in its absence. Why isn’t there DNA connecting the defendant to the crime? may be a common thought for a juror in such a case, meaning the prosecution has to work harder to prove their case.
In this case, Shirley was convicted. The prosecution presented additional evidence linking the defendant to the crime besides DNA, such as witness testimony, and the jury found the defendant’s guilt proved beyond a reasonable doubt.
However, Shirley’s case was not the first to result in the exclusion of Y-STR evidence in a sexual assault case. See U.S. v. Kootswatewa, 2016 WL 808663 (D. Arizona 2016). Nor may be it the last. The issue presented by Y-STR evidence for Native Americans is one that the scientific community may help us solve, whether that involves accounting for population substructure in statistical analysis,  expanding Y-chromosome databases to include more robust population samples, increasing the numbers of Y-STR markers or developing full sequencing of the Y-chromosome so as to make matches more rare, or questioning the premises on which our reliance on racial categorization in the case of the Y-chromosome rest.
Even more pioneering would be for scientists to improve the processes of obtaining autosomal DNA from mixed samples of victim and perpetrator DNA. This would involve perfecting and expanding the use of single cell analysis. For the present, single cell forensic analysis remains uncommon because of its expense but ultimately, such technology has the potential to decrease our reliance on the Y-chromosome for forensic analysis.
Nicole Bronnimann, JD class of 2018, Stanford Law School
 CODIS markers are a set of 13 autosomal short-tandem-repeat (STR) markers. “CODIS” refers to the Combined DNA Index System, a database maintained by the FBI that houses profiles derived from these markers.
 A Daubert hearing, so-called after the case Daubert v. Merrel Dow Pharmaceuticals, 509 U.S. 579 (1993), is a pre-trial hearing held to determine the admissibility of expert witnesses’ testimony during a trial in federal court. Under Daubert, a judge acts as “gatekeeper” of expert testimony and must determine whether it is relevant to the case and based on a reliable scientific methodology.
 The government originally intended to introduce odds of 1 in 35 based off the older ABI population database. The revised 1 in 76 figure was based on using the expanded US-YSTR database.
 Michael Hammer & Alan Redd, Forensic Applications of Y-Chromosome STRs and SNPs, https://www.ncjrs.gov/pdffiles1/nij/grants/211979.pdf
 “The results indicate that Native American populations have lower levels of Y chromosome diversity than other U.S. ethnic groups. This is reflected in a higher percentage of shared haplotypes and higher random match probabilities.”
 State v. Calleia, 414 N.J.Super. 125 (2010)
 This phenomenon has been coined the “CSI effect,” in reference to the popular television show CSI, which dramatizes forensic investigation. A 2008 study on the so-called “CSI effect” found that viewers of CSI were more likely to expect forensic scientific evidence from the prosecution but were not significantly more or less likely to acquit defendants without scientific evidence. Only four of thirteen scenarios presented to participants found a somewhat significant variation between CSI-viewers and non-CSI viewers. Interestingly, one of these was the rape scenario; CSI viewers were slightly less likely to convict absent scientific evidence. Donald E. Shelton, The ‘CSI Effect’: Does It Really Exist?, 259 Nat’l Inst. Justice, Mar. 2008, at 1, 5.
 One promising sign is that in 2015, the National Institute of Justice awarded a grant to the University of Washington to address “Population Genetic Issues for Forensic Y-Chromosome Markers,” such as “the effects of population structure on match probability calculations.”
 Allan Jamieson & Scott Bader, A Guide to Forensic DNA Profiling 390 (2016). For recent studies involving improving single cell analysis, see S. Brück et al., Single Cells for Forensic DNA Analysis—From Evidence Material to Test Tube, 56 J. Forensic Sci. (Jan. 2011), available at https://www.ncbi.nlm.nih.gov/pubmed/21198592; T. Geng, R. Novak, & RA Mathies, Single-cell Forensic Short Tandem Repeat Typing within Microfluidic Droplets, 86 Analytical Chemistry (Jan. 7, 2014), available at https://www.ncbi.nlm.nih.gov/pubmed/24266330.