Whole Genome Sequencing at the Legal Threshold

Perspectives of Dr. Kelley Harris, testifying expert in the case of Rex Heuermann

BY LERON VANDSBURGER

In July 2023, Rex Heuermann, a New York architect, was arrested and charged with seven murders between 1993 and 2010 in the Gilgo Beach serial killing case. Prosecutors’ evidence against Heuermann involves phone records, internet activity, and a type of DNA analysis called whole genome sequencing, or WGS, that is new to the courtroom. 

Dr. Kelley Harris, an associate professor in the Department of Genome Sciences at the University of Washington and an affiliate faculty member in Computational Biology at the Fred Hutchinson Cancer Center, was recently called to testify as an expert witness for the prosecution in Heuermann’s New York Frye hearing, a type of pre-trial proceeding used to determine the admissibility of scientific evidence or expert testimony. At least in part due to Dr. Harris’ testimony during the hearing, the judge admitted WGS evidence for confirming the identity of the defendant, a first in the United States. 

Washington’s Current Framework

Under Washington law, DNA evidence is admissible if it satisfies the Frye general acceptance test¹1 The Frye general acceptance standard governs the admissibility of scientific expert evidence by requiring that the underlying method or principle be generally accepted as reliable within the relevant scientific community. It originates from the 1923 case Frye v. United States, in which a court excluded early lie-detector evidence because the technique lacked such acceptance. Frye became the dominant U.S. standard for decades and remains in use in some states today, particularly for assessing novel scientific evidence, being replaced in some jurisdictions by the Daubert standard that originated in Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993). and the ER 702 standards for expert testimony. The Washington Supreme Court has repeatedly emphasized two things: general acceptance among relevant scientists, not necessarily unanimity; and reliability of methodology, as distinct from case-specific questions of chain of custody or laboratory error.

The long-established standard for forensic DNA testing in Washington is called short tandem repeat (STR) analysis. STR analysis targets short sequences of DNA, typically two to six base pairs in length, that repeat at specific locations in a person’s genome. Because the number of repeats at each location varies greatly between individuals, analyzing a standard panel of 13–20 STR locations produces a DNA profile with a high degree of individuation. STR analysis has been used in forensic casework since the late 20th century, is the basis of the FBI’s CODIS (Combined DNA Index System) database, and has been repeatedly upheld as admissible under the Frye standard. Its reliability, reproducibility, and general acceptance among forensic scientists established it as the benchmark against which newer technologies—such as whole genome sequencing—are currently being judged.

In contrast, WGS reads a person’s full DNA sequence—three billion base pairs—to locate genetic variations. To be admissible in Washington courts, as in New York, WGS needs to satisfy the same criteria as STR analysis. The question for Washington courts will be whether the technology has moved far enough from research labs into mainstream scientific practice to be considered “generally accepted.” 

Inside the New York Hearing

According to Dr. Harris, the New York Frye proceedings revolved around two themes: whether WGS techniques have reached the threshold of general acceptance for use as evidence at criminal trials, and more technically, how WGS actually works. Dr. Harris’s position is that WGS is routinely used in genetic sciences and has become commoditized, and that it is technologically mature enough to join STR as evidence in courts.

Leron Vandsburger (LV): Washington and New York use the Frye standard, where new science is admitted only if it’s generally accepted in the relevant community. Do you feel that whole genome sequencing is bleeding edge right now or do you feel like it’s accepted?

Dr. Kelley Harris (KH): I testified and I believe that whole genome sequencing is no longer bleeding edge… . [T]he Human Genome Project²2 The Human Genome Project was a large international collaboration that produced the first human genome sequence, bringing WGS to bear on human genetics for the first time. was around the year 2000. At that point there had been one human genome, and now there [are] several hundred thousand human genomes that have been sequenced … I can just send a sample down the hall and get them to produce a whole genome for about $500. And I think once something is that standard to produce in a core facility, it’s hard to call it bleeding edge.

TESTING THE SCIENCE

At the New York Frye hearing, the defense challenged the lack of forensic lab accreditation for WGS, the reliability of population datasets such as the “1000 Genomes Project,”³3 The 1000 Genome Project was an international collaboration that published about 2,500 whole genome sequences between 2010 and 2015, creating a reference panel of genetic diversity from several European, African, East Asian, South Asian, and Amerindian populations. and the trustworthiness of the software pipelines (a software pipeline can be described as a computer program that, in this case, takes input data and transforms it into a different kind of processed data that is more scientifically interpretable). 

LV: During the proceedings, the defense objected on several grounds: whether the lab was authorized to run the tests, whether population datasets like 1000 Genomes were reliable, and whether the software tools could be trusted. Which of those challenges struck you as the most serious?

KH: I wasn’t asked to address them directly. I could tell the questions were organized around some of these points. I think as far as the accreditation question goes, that was the farthest from my expertise, because I’m not really a forensic scientist. But it seems like accreditation just would not be in play at this point, because a lab is [only] accredited to do stuff that’s currently allowed in court … [T]hat was not this, at least prior to the judge’s recent Frye ruling. And so why would they be accredited?

While the wet-lab⁴4 “Wet-lab” is a colloquial name for experimental science involving laboratory chemicals. In contrast, “dry lab” denotes computational data analysis. Both wet lab and computation/dry lab are required for forensic genetics—wet lab is the first step and dry lab analysis is the second step. sequencing steps went largely unchallenged, both sides presented arguments centered on bioinformatics: aligning genome segments derived from degraded samples to the human reference genome, filtering errors, and calculating likelihood ratios.

LV: Could you walk us through the workflow of a forensic whole genome sequencing
case—from testing a rootless hair, as was done in New York, all the way to generating a report? And where are the weak points that tend to draw courtroom challenges?

KH: So a rootless hair is, you know, a thing that was once alive and is now dead. The thing that makes that challenging is once cells die, they don’t have their nice machinery to repair damage to DNA. For our cells that are alive, a lot of what they do, what their metabolic energy is devoted to, is maintaining the DNA, making sure it isn’t broken, repairing it if it gets broken, checking for the sort of damage that could cause dysfunction or cancer. And so, all of that is going on in living cells. If you want to sequence living cells, you have what’s called high molecular weight DNA, meaning long, intact chromosomes. That’s what a cell needs to be alive and [what makes it] easy to sequence. But a rootless hair is not living cells anymore, and the longer it’s been sitting somewhere and the harsher conditions it’s been in, the more the DNA gets broken up into pieces.

FROM PIXELS TO PICTURES

As noted earlier, STR analysis reads short sequences of DNA that repeat at certain spots in a genome, whereas WGS reads the entire genome. Dr. Harris likened STR profiles to grainy surveillance photos, while WGS provides a high-resolution picture—albeit one that may be blotchy if samples are degraded.

LV: What unique value would you say whole genome sequencing brings to casework or the courtroom? And where do you see its limitations in practice?

KH: Well, I think to see the value it brings, imagine trying to recognize the face of a suspect. … STRs are a little bit more like if you had an extremely pixelated image, [whereas WGS is like] if you had two high-resolution photos. Most of us can compare [two] photos and feel really confident saying this is definitely the same person or this is definitely not. … [A] whole genome has, give or take, 3 billion base pairs of information, 6 billion if you count the copy from mom and the copy from dad. And STRs … are just a lot fewer pixels to work with. So you’re not going to be as confident.

LV: Could you explain what the two high-resolution photos are here? Are you describing the human reference genome?

KH: The human reference genome is the first human genome that got sequenced in the year 2000, or an updated variant, an updated version of that. The two “photos” are the samples that are tested to assess probability of a match. One sample could be taken from a crime scene and the other could be taken from a suspect in custody. For example, in the New York case, one sample was from the defendant and another was from rootless hairs collected from one of the bodies. The reference genome is used as a template for assessing the probability of a match between the two samples.

As long as the fragments you have are … sufficiently long, you can scan through the whole human genome and say, “OK, this spot on chromosome 16 … is the most similar to this fragment of DNA, and this is … the place in the puzzle where it goes.” [This is done with a] very standard software program called [Burrows-Wheeler Aligner or] BWA.

You take all the fragments from your samples [and] you figure out where they go with respect to the reference human genome. And, of course, these fragments aren’t going to match exactly, but they usually will overlap with one another. The better quality your sample is, the more of what’s called “coverage” you have. So, when you’re getting a genome from a living person and you want it to be high quality, usually you’ll go for what’s called 30x coverage, meaning every position in the reference genome is covered by an average of 30 different short fragments.

LV: Because … once you’re confident where these pieces of DNA go in the genome, you want to know, how are they different from the reference human?

KH: Two people who are different are usually different in about one out of every thousand base pairs, give or take. What you’re trying to identify is which one in a thousand base pairs differ between the samples you have.

If you have a highly degraded, low input sample, you probably won’t have the luxury of 30x coverage. Instead, you’ll have a bit of DNA here and there and it’ll have some differences from the reference sample … .
[But] you won’t be able to tell just by law of large numbers which genetic differences look real and which ones look fake. So then you need to use other principles down the line to sort out the errors from the true variants. And that’s one of the trickier aspects of [analyzing whole genome sequencing data from a degraded sample] compared to whole genome sequencing of living cells or other high-quality samples.

LV: So it sounds to me that, in the case of a rootless hair or some other degraded sample, you’re somewhere in between—in terms of the quality of your data that you can use in court—the pixelated security camera and the high-resolution image.

KH: The pixelation [with WGS] is different. With STR, you only get to look at specific places on the genome, which is like picking specific pixels in the face, in advance, without knowing if that place in your sample will be degraded or not or will have rare genetic variants or not. What you have with a degraded sample in WGS is like starting with a high-quality photograph that shows everything [except for] random bits of white or red or other colors … that will make your photo look a little funny, but you’ll still have a lot of information left over. [P]robably more information than if you just sort of said, “OK, I’m picking 20 pixels and I’m going to get them really well.”

WGS IN ACTION

LV: Once you have the WGS data, how is it used to identify an individual?

KH: … [First, you] identify in each of your sequencing reads where it matches … the sample collected directly, such as from a defendant, and where it doesn’t. The next step involves a more bespoke program called [Identity-By-Descent GEM or] IBDGEM, which was scrutinized in the Frye hearing in New York. This takes a panel of high-quality human genome sequences that should be randomly drawn from a large representative population. This panel is used to identify places in the genome where people tend to be variable.

LV: How does the panel help?

KH: If you and I compare our DNA … we’re going to probably be different in one out of 1,000 positions, but [does] that indicate how a third person might differ from each of us? Basically, [yes,] the third person is much more likely to differ at positions where we’ve already found a difference between you and I. And the reason for that is mutations that happened a long time ago in some human population [and] that reached a high frequency are now carried by large parts of the [the modern-day] human population. Say for example, the mutation that caused variation between blue and brown eyes. That’s a high-frequency variant where a bunch of people have one thing and a bunch of people have another thing. Those differences occur in consistent locations on the genome.

LV: Can you help me understand why this is important for identifying a person?

KH: So if you see that the reference genome has the brown-eyed variant and then we found a read in our sample that has the blue-eyed variant, that’s a variant we’ve seen a lot and we can recognize it. We’re more likely to believe that’s real versus a variant that we’ve never seen before in a large panel of people. It still could be real, but it’s a lot more likely to be error. What the panel lets us do is focus on all the sites where we know real variation exists and decide based on those more trustworthy sites. It’s like we’re focusing on informative parts of a human face, like the shapes of eyes and noses that are useful for identification, rather than other parts, like a random patch on the forehead that looks more or less the same in all people. The advantage of this [WGS] approach is that it allows high-quality regions of the genome to be used for identification, rather than choosing regions in advance before knowing if the data will be usable [as in STR analysis]. … [N]ot to mention that the number of high-quality regions [with WGS] isn’t limited in advance, like in STR panels, to a few specific places.

DETERMINING KINSHIP

LV: At least part of the DNA evidence in the New York trial concerned identifying the wife or daughter of the defendant. How effective is WGS at identifying kinship?

KH: It’s very, very well established to be able to say, take two high-quality DNA sequences and be able to tell if they are parent and child or siblings or first or second cousins or even, I mean, beyond second cousin. 

With STRs, it’s much harder to get accurate kinship estimation. … [T]his was something that made a big impression on me when I was in grad school. A lab mate of mine, who did more forensic work, was talking about this law in California, at least at the time, called familial search, which said that, if based on STR evidence, you think that a sample at a crime scene is likely to be a relative of someone who’s in the CODIS database, you can then investigate the family of the person in the CODIS database. And what my lab mate was doing was showing that with STRs, there is a high chance of saying that you think these two people are brothers or cousins, when that’s not true.

LV: So there’s a higher risk of a false positive with STR than there is with whole genome sequencing? 

KH: Yes. When it comes to deciding whether two people are a certain type of close relative, definitely.⁵5 Rori v. Rohlfs, Erin Murphy, et. al, “The Influence of Relatives on the Efficiency and Error Rate of Familial Searching,” PLos One, Aug. 14, 2013, available at https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0070495.

LV: Is there a consensus that whole genome sequencing is good for the courtroom? Is it courtroom ready or is there still division amongst the research community or even the forensic community that you’re aware of? 

KH: I mean, I think good for the courtroom isn’t just a scientific question, it’s a societal question. So I think scientists would say things like … it’s much better for telling if people are close relatives and much, much better for telling what ethnicity they are. You can get more certainty about if two samples are the same person or not.

Why It Matters for Washington

LV: Washington law requires long-term preservation of DNA evidence in serious cases. With whole genome sequencing, that means not just a profile, but raw sequence data and metadata. How should labs and prosecutors handle that responsibility?

KH: That’s a great question because there’s a practical dimension. A whole genome sequence, depending exactly how you store the information, takes megabytes to gigabytes of storage. … So, I know academic labs that will say, oh, it’s cheaper in the long run to delete some data and possibly have to sequence it again later because the sequencing has gotten so cheap that it’s cheaper than cloud storage. I personally have mixed feelings about whether genome data should be stored in a publicly searchable database, like what law enforcement would use, at least because privacy would be difficult to maintain. 

LV: Why would you have mixed feelings?

KH: There definitely are some cases where it could cause harm. Things like the BRCA gene, which, if you have a specific variant, it says something about your lifetime cancer risk. You could sequence someone’s genome for forensic identification purposes and figure out that they have this high lifetime cancer risk. And, you know, maybe telling them about that is a favor. Maybe … that is something they consciously didn’t want to know.

I think there are ways you could deal with this. You could do whole genome sequencing and then mask out regions that you know to be sensitive. So, for instance, only 2 percent of the genome is [made up of] genes that code for protein, which is where most of the disease risk information has been identified to date. For most genes, we don’t know if their variation has anything to do with disease. If you wanted to be really conservative and mask out all of the protein-coding genome, you would still have 98 percent left over for identification, which would be just as good for forensic purposes.

Conclusion

The New York Frye ruling signals that whole genome sequencing is reaching the general acceptance threshold. As Dr. Harris put it, the move from STRs to WGS is like moving from a pixelated image to a high-resolution photograph. Whether Washington courts will agree that the picture is clear enough for juries remains to be seen. 

---

SIDEBAR

Why It Matters for Washington

Washington and New York both use the Frye test, so the New York case is an instructive preview. Three issues stand out for Washington courts and lawyers:

• Expert qualifications. Courts will need to decide whether expertise must lie in wet-lab genetics, or whether computational biologists—those trained in bioinformatics and statistical genetics—are the more relevant experts for interpreting WGS evidence.
• Data retention. Washington requires long-term preservation of DNA evidence in serious cases. With WGS, this means storing not only a genetic profile but raw sequencing files—potentially gigabytes per case. Dr. Harris noted that academic labs sometimes delete data rather than pay for cloud storage, raising practical and legal challenges.
• Privacy concerns. Whole genomes contain medically sensitive information, far beyond identification. Dr. Harris suggested that forensic labs might redact disease-related regions, preserving discriminatory power while protecting privacy. One risk, however, is that as science progresses, additional disease-related regions may be revealed in DNA data that has already been published or stored.

NOTES

1. The Frye general acceptance standard governs the admissibility of scientific expert evidence by requiring that the underlying method or principle be generally accepted as reliable within the relevant scientific community. It originates from the 1923 case Frye v. United States, in which a court excluded early lie-detector evidence because the technique lacked such acceptance. Frye became the dominant U.S. standard for decades and remains in use in some states today, particularly for assessing novel scientific evidence, being replaced in some jurisdictions by the Daubert standard that originated in Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993).

2. The Human Genome Project was a large international collaboration that produced the first human genome sequence, bringing WGS to bear on human genetics for the first time.

3. The 1000 Genome Project was an international collaboration that published about 2,500 whole genome sequences between 2010 and 2015, creating a reference panel of genetic diversity from several European, African, East Asian, South Asian, and Amerindian populations.

4. “Wet-lab” is a colloquial name for experimental science involving laboratory chemicals. In contrast, “dry lab” denotes computational data analysis. Both wet lab and computation/dry lab are required for forensic genetics—wet lab is the first step and dry lab analysis is the second step.

5. Rori v. Rohlfs, Erin Murphy, et. al, “The Influence of Relatives on the Efficiency and Error Rate of Familial Searching,” PLos One, Aug. 14, 2013, available at https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0070495.