OPINION OF THE COURT
Defendant is charged with predatory sexual assault (Penal Law § 130.95 [1] [a]) and burglary in the first degree (Penal Law § 140.30 [1]). During pretrial proceedings, the court learned that the People intended to introduce DNA evidence as part of their case-in-chief. Defense counsel was provided with copies of the DNA report. Following his review of the report, defense counsel raised the question of the admissibility of the report as well as testimony relating to the report and thereafter sought a Frye hearing (see Frye v United States, 293 F 1013 [DC Cir 1923]). After consideration of written submissions, the court concluded that it would conduct a Frye hearing.
The specific issue here is the admissibility of the forensic DNA testing results performed on a number of items of evidence discovered at the crime scene (the victim’s house) and on cuttings from a dried red-stained area of a sock found in defendant’s bedroom. DNA testing on those items was conducted by the Erie County Central Police Services Forensic Labora
The parties have provided the court with voluminous submissions in support of their respective positions. Those submissions include references to relevant cases, articles from scientific journals, and expert affidavits. The court recognizes that the science behind DNA analysis and statistical probabilities is complex. This court, however, previously rendered a decision involving the admissibility of a DNA “kit” (see People v Borden, decision attached to People’s mem of law, dated Dec. 23, 2015) and therefore is familiar generally with the scientific principles at issue.
Before summarizing the testimony from the Frye hearing, the court believes that some background discussion of DNA analysis and interpretation is necessary. Rather than attempt to reinvent the wheel, the court has taken the liberty of quoting at length from Judge Michael Coccoma’s recent decision involving a similar probabilistic genotyping program:
“DNA identification is a powerful forensic tool for solving and preventing crime. Two common sources of data ambiguity in biological evidence are DNA mixtures from multiple contributors and low-template (evidence samples below the threshold) DNA. Although some American laboratories are moving to quantitative modeling of DNA mixture data, most still use combined probability of inclusion (CPI) or combined likelihood ratio (CLR), using the qualitative Boolean logic of all-or-none allele (the number of repeated words) events. Both approaches apply thresholds to the DNA data that cut off quantitative information. Their analysts subjectively apply these analytical or stochastic thresholds manually to data peaks to decide whether or not they believe the evidence peak represents an allele in the genetic material. But the more complex data that has mixtures or low-template DNA limits the applicability of such qualitative procedures.
*180 “Computer interpretation methods use more of the quantitative short tandem repeat (STR) peak height data rather than thresholds and have been used for over 20 years. Computers offer three principal advantages in the interpretation process: (1) productivity—eliminates the often time-consuming human review of cases that are impossible to solve, (2) information—human review typically makes simplifying assumptions that can discard considerable identification information containing DNA evidence whereas a computer can use a statistical model to fully examine the quantitative peak height data, and (3) objectivity—human mixture interpretation methods sometimes use the suspect genotype (pair of allele) to help infer or report results whereas a mathematically programmed computer can infer a genotype . . . from the evidence data without using any suspect information and then afterward compute a match likelihood ratio (LR) statistic from this genotype. “Probabilistic genotypes have been recognized by regulatory bodies such as the Scientific Working Group on DNA Analysis Methods [SWGDAM] in its 2010 Interpretation Guidelines for Autosomal STR Typing by Forensic DNA Testing Laboratories and the American National Standards Institute [ANSI] in the 2011 article Data Format for the Interchange of Fingerprint, Facial & Other Biometric Information as a valid approach to DNA Interpretation and reporting. There are two probabilistic approaches:
“(1) semi-continuous—information is determined from the allele present—peak heights are not considered, and
“(2) fully continuous—incorporation of biological parameters.” (People v Wakefield, 47 Mise 3d 850, 852-854 [Sup Ct, Schenectady County 2015] [footnotes omitted].)2
The Frye Hearing
The Frye hearing was held on January 11, 14, and 21, 2016. The People’s only witness was Dr. John Simich, the director of
After the PCR/STR process is complete, an instrument called a Genetic Analyzer reads the “various amplified fragments of DNA, and then that translates it into something that a human can see which is the electropherogram” (1-24). Once the Lab generates a DNA profile from the sample, that profile is compared to an individual—victim, suspect, or elimination sample—“to determine if they are the source of that DNA” (1-27). An electropherogram is a printout of the “graph of the various DNA types that were identified at each of the genetic markers” (1-27). Exclusion can be made by visual comparison of the electropherograms. If there is no exclusion, the Lab needs “to determine the weight of the evidence” (1-28).
The electrophoresis step produces a chart, which gives the value of the DNA marker at a certain point and the “strength” of the signal (a measure of how many of the DNA molecules were examined at a particular value) (1-32). In July 2015, the Lab began using STRmix to make this calculation. STRmix allows the Lab to report the results as a likelihood ratio.
STRmix uses “continuous probability genotyping” software (1-34). That software uses information that has been available for “years” (1-34). STRmix was recommended by different scientific organizations as “the best way to perform reporting of the weight of the DNA evidence” (1-34). A “continuous”
Dr. Simich testified about various scientific organizations that review the software programs: SWGDAM, National Institute of Standards and Technology, and International Society of Forensic Genetics. According to his undisputed testimony, all three have recommended STRmix. STRmix addresses the problem of mixed samples, that is, DNA with more than two contributors. STRmix “is able to break [the DNA sample] down into its component mixtures” (1-38).
Dr. Simich discussed the concept of “probabilistic genotyping,”
Dr. Simich was familiar with some of the mathematical analysis methods and principles used in the deconvolution process, for example, the MCMC (Markov Chain Monte Carlo) model, and the Metropolis-Hastings algorithm. MCMC is a standard statistical modeling process. STRmix also employs Bayes’ theorem, which is a general scientific principle of the likelihood ratio. Bayes’ theorem was developed in the early 1700s and has been used for centuries in various scientific disciplines without controversy (1-45).
After reviewing the STRmix software, Dr. Simich concluded that the science behind it was generally accepted within the forensic lab community. He further concluded that the software was reliable, based upon his review of peer-reviewed journals
Dr. Simich testified that he had considered other software programs, including Forensic Statistical Tool (FST) and TrueAl-lele. He concluded that “STRmix is accurate and reliable and can be utilized to generate likelihood ratios for mixture decon-volutions” (1-59).
Dr. Simich testified about the approval of STRmix by the New York State Commission on Forensic Science (the Commission). The Commission is the “organization that actually will evaluate and grant . . . New York State accreditation” (1-60). The DNA Subcommittee, made up of experts in various scientific disciplines related to DNA analysis, evaluates the DNA aspect of forensic labs in New York. Dr. Simich appeared before the Subcommittee three times and presented his validation studies; Dr. Buckleton, one of the creators of STRmix, also appeared before the Subcommittee. In May 2015, the Subcommittee voted 4-0 (with one member abstaining) on a binding recommendation to the Commission to allow the Lab to use STRmix for casework analysis (see minutes of the meeting of the DNA Subcommittee approving STRmix, People’s exhibit 5; letter from the Subcommittee to the Chair of the Commission with the binding recommendation, People’s exhibit 6; 1-63). The DNA Subcommittee also issued a general recommendation on the use of probabilistic genotyping software (People’s exhibit 7).
Dr. Simich testified without any contradiction that STRmix is used in other labs in Australia and New Zealand, and by United States Army Criminal Investigation Laboratory, the California Department of Justice, and the FBI. He was aware that some labs use FST (New York City’s Office of the Chief Medical Examiner, or OCME) or TrueAllele (Virginia, some labs in Pennsylvania). He believed that other labs in New York
On cross-examination, Dr. Simich explained why he selected STRmix for the Lab over other software programs. He admitted that there were other acceptable methods to perform forensic statistics (1-79) and explained the differences among the various accepted methods (1-80). He opined that, in this case, the DNA profile was “15,000 times more probable if the sample originated from [defendant] and three unknown unrelated individuals because it’s a four-person mixture rather than if it originated from four unknown unrelated individuals in the U.S. population” (1-88).
Dr. Simich was asked about the “reproducibility” of the results reached by STRmix, in the context of his testimony that every time a sample was analyzed there would be a different likelihood ratio (1-115). He testified that variations among the different results were not “statistically significant” (1-115). The People rested, relying solely on the testimony of Dr. Si-mich and the accompanying exhibits.
Defendant presented the testimony of Dr. Gary Skuse, who teaches biological sciences at the Rochester Institute of Technology, with a specialization in a field called bioinformat-ics, which involves the interaction between biology and computers (III-4). His experience with DNA forensics consists of working “with criminal defense attorneys primarily helping them understand the processes that go into using DNA in criminal cases and help them interpreting the results” (III-6-7). He testified in the Wakefield case about the “way the DNA was isolated, the way the laboratory interpreted the DNA results” (III-7). With respect to STRmix, he reviewed various articles as well as material from the company, and the protocols established by the Lab. He criticized the amount of “human intervention and human judgment” involved in setting up the software (III-9). He was “interested that a laboratory acknowledged that the software itself gives different answers every time it’s run” (III-9). He discussed the notion of “objective science” and opined that STRmix was not acceptable in accordance with general scientific principles (IIX-10).
On cross-examination, Dr. Skuse conceded that he had no training in forensic DNA analysis and that he had never been to the Lab. He apparently has visited the Monroe County Forensic Laboratory on rare and sporadic occasions. He had never used the STRmix program (III-50).
Defendant rested, and the parties provided the court with further written submissions; oral argument was held on February 25, 2016.
The Law
The Frye Standard
Frye established the general proposition that scientific expert witnesses are only permitted to give opinion evidence when their testimony is based upon scientific principle or discovery that has passed the mere experimental stage and become demonstrable scientific knowledge generally accepted as valid within the relevant scientific community. Even though the federal system, and a number of states, have moved away from the Frye standard to one embracing a more hands-on gatekeeper function for the trial judge (see Daubert v Merrell Dow Pharmaceuticals, Inc., 509 US 579 [1993]), New York continues to follow the Frye approach. Initially, the proponent of the evidence provides case citations indicating that other courts have already addressed the issue and approved its admissibility. Such cases can come from any competent court and jurisdiction, whether in the United States or elsewhere. The proponent will also typically reference learned scientific treatises, published papers, books or other learned writings demonstrating acceptance of the scientific principle within the applicable scientific community (see People v Hughes, 59 NY2d 523 [1983]). When there are no such published materials, or when they are sparse or conflicting, the proponent must augment his proffer by expert testimony at a pretrial Frye hearing.
At such a hearing, the proponent bears the burden of proof. The standard that the proponent must meet does not seem to have been clearly delineated by New York courts. At oral argument, the People indicated they were unable to find any specific authority on this issue (tr of Feb. 25, 2016 at 21). The People rejected defendant’s suggestion that the appropriate standard is “beyond a reasonable doubt,” and argued that the more appropriate standard is “preponderance of the
The proponent of the evidence must establish that the scientific principles and techniques he advocates, when properly performed, generate consistent results accepted generally as reliable within the relevant scientific community (see People v Wesley,
“a court assessing the admissibility of evidence under Frye is not charged with deciding the validity of novel scientific procedures. It would hardly be sensible to assign that task to the judiciary, most of which is . . . patently unqualified to perform the task as is this court. Judges should be ‘counting scientists’ votes,’ and not ‘verifying the soundness of a scientific conclusion’ ” (People v Collins,49 Misc 3d 595 , 603 [Sup Ct, Kings County 2015], quoting Parker v Mobil Oil Corp.,7 NY3d 434 , 446-447 [2006], quoting People v Wesley,83 NY2d 417 , 439 [1994, Kaye, Ch. J., concurring]).
While the issue of the admissibility of DNA has long since been resolved, new issues have arisen regarding the interpretation of the results of DNA testing. There are numerous cases in New York regarding software programs that interpret DNA results, although none involve STRmix. Although the scientific principles underlying the STRmix program are similar to the principles if not identical to the programs that have been considered, and almost universally accepted by courts in New York, this case concerns the first judicial review, as far as this court is aware, of STRmix in New York. For that reason, the court decided to conduct a Frye hearing.
Defendant’s primary objection to the admissibility of the results of the DNA testing, advanced in the memorandum of law received on February 16, 2016, is that the relevant scientific community for purposes of deciding whether STRmix is accepted is “an insular community of professionals whose careers and livelihoods focus on the prosecution of criminal cases and upon the ‘discovery’ of inculpatory evidence.” Defendant urges this court to “take the warnings of Frye and Leone to heart.”
First, Dr. Simich was thoroughly familiar with the application of the STRmix software. While he could not expound on the underlying mathematics, his lab conducted validation studies (People’s exhibit 2) and he reviewed numerous articles regarding the software. As noted in Muhammad, the mathematical models are themselves non-controversial and have been widely used in fields such as weather forecasting, computational biology, linguistics, genetics, engineering, physics, aeronautics, finance, and social sciences. As the director of a forensics lab, Dr. Simich is well-qualified to critique software
With all due respect to Dr. Skuse and not in any way to denigrate the intelligence and experience that he so obviously possesses, this court does not believe that he can be considered as an expert in the field of forensic DNA analysis in general or on the specific topic of the scientific acceptance of probabilistic genotyping as utilized.
Although the court would have preferred to hear from other experts in the relevant scientific community,
The role of the Commission and the DNA Subcommittee in this court’s decision, while not dispositive, certainly cannot be discounted. In 1994, the legislature created the Commission (Executive Law art 49-B) to “develop minimum standards and a program of accreditation for all forensic laboratories in New York state . . . and approval of forensic laboratories for the performance of specific forensic methodologies” (Executive Law § 995-b [1] [emphasis added]). To achieve its mission, the Commission is designed to “ensure that forensic analyses, including forensic DNA testing, are performed in accordance with the
The record establishes that the DNA Subcommittee met on May 29, 2015, and voted unanimously (4-0, with one abstention) “to issue a binding recommendation to the [Commission] that the use of STRmix by [the Lab] be approved for forensic casework” (People’s exhibits 5, 6). The DNA Subcommittee met again on August 14, 2015 to discuss a June presentation by Dr. Michael Coble at the National College for Forensic Science Litigation on the topic of “Software Systems for Interpreting Low Level Samples and Complex Mixtures” (People’s exhibit 7). The letter sent to the Commission following that meeting is informative. It states in relevant part:
“The Subcommittee indicated that the presentation addresses some of the challenges involved in handling complex mixtures and some of the advantages of using probabilistic models to aid in interpretation. Subcommittee members agreed that Dr. Coble’s presentation substantiates the need to embrace new technologies, especially when challenging mixtures require additional analytical methods to assist in their interpretation. The presentation addressed the benefits of using these software tools and revealed that those laboratories not using deconvolution software tools are at a disadvantage. Members noted that interpretations without the software were ‘all over the road.’
“In sum the Subcommittee members expressed that they are pleased that New York State labs are moving forward to validate these software systems. All Subcommittee members were resolute that the development and use of these software tools is a significant advancement and will greatly assist laboratories in the analysis, interpretation and reporting of DNA mixtures.”
The Commission voted on June 19, 2015 to approve the binding recommendation of the DNA Subcommittee.
In addition to the votes of the bodies appointed by state law to consider the use of various forensic DNA methodologies, the court is also faced with the almost unanimous approval of the courts that have considered similar software programs. Of course, there is no reported case in New York regarding the admissibility of the STRmix software program, but the court is persuaded by the analysis of those courts that have reviewed similar programs (see People v Megnath,
Only two of the New York cases are worth discussing in detail. The first is People v Wakefield, which involved TrueAl-lele. TrueAllele, like STRmix, involves probabilistic genotyp-ing. The court recognizes that STRmix and TrueAllele are not identical and in fact defendant argues that the differences between the two are so significant that the Wakefield decision cannot be used to support a ruling in favor of STRmix. The court also recognizes that TrueAllele is a competitor of STRmix.
In any event, the court in Wakefield discussed at length the role of the Commission and the DNA Subcommittee, both of which voted to approve TrueAllele, and concluded that “approval” by the Commission and the DNA Subcommittee “clearly constitutes ‘general acceptance’ ” (Wakefield,
People v Collins (
Additionally, Justice Dwyer’s criticism in Collins that validation studies are not “conclusive” because they are only
The Collins decision was criticized in People v Carter (
The last submission by defendant, received by the court on February 16, 2016, perhaps in an attempt to lighten the tone of these otherwise dreary briefs, contains the following quotation from Ernest Rutherford: “If your experiment needs a statistician, you need a better experiment.”
“if you run the program twice with the same input . . . [b]ecause the MCMC process is a random process, and it generates a random number to begin the process every time you do the analysis. So you will get a different likelihood ratio every time you . . . put the same data in” (1-114-115).
In a broader and more general sense, Dr. Skuse countered that “traditional scientists” will do an experiment three times and if the results are “close, you report them ... or you figure out what’s wrong and try it again” (III-12).
The court finds that whether the procedure is performed once as recommended, or three times (or more) as Dr. Skuse seems to suggest, it does not affect the question this court is called upon to decide, namely, general acceptance of the probabilistic genotyping procedure within the relevant scientific community. Of course, this court would expect that the statement referred to above would add another arrow in the quiver of defense counsel that would be used to undermine the STRmix results when the issue is presented to the trial jury, but it does not affect the issue of the general acceptance of STRmix within the relevant scientific community.
People v Muhammad is the only other reported case in the country regarding the admissibility of STRmix. There, the court concluded as a preliminary matter that “statistical evaluation of the [DNA] analysis’s results is a matter of evidentiary weight, not admissibility.” {People v Muhammad, slip op at 1.) Thus, the court’s determination of admissibility falls into the category of dicta. Nonetheless, the court reached several conclusions, which are persuasive insofar as this court is faced with identical issues. First, the Muhammad court found that STRmix “received adequate validity testing.” (People v Muhammad, slip op at 2.) Indeed, Dr. Buckleton testified in Muhammad and it was anticipated, based on preliminary representations made to this court by the People, that he would testify here. His testimony could have resolved several questions
In conclusion, the court would like to commend the parties for their passionate and thorough advocacy on behalf of their respect positions. Contrary to the statement in the last submission by defense counsel, received by the court on February 16, 2016, the court does not expect its decision to be “unassailable throughout the state for the remainder of eternity.” The court is aware that there is a case pending in Erie County, with a Frye hearing scheduled, on this very issue. As other laboratories throughout New York and the country adopt STRmix, courts will deal with the same questions presented here. At oral argument the parties agreed that there were other programs similar to STRmix being used in labs. Dr. Simich testified that he considered several programs before settling on STRmix. This court’s decision is based on the testimony adduced at the hearing, the accompanying exhibits, and the relevant case law. It may be among the first words in New York courts on the admissibility of STRmix, but the court certainly does not expect it to be the last.
Notes
. According to the STRmix website (http://strmix.esr.cri.nz), “STRmix is a breakthrough for forensic analysts as it can assist investigations using DNA evidence that was previously considered too complex to interpret. The software has been developed by New Zealand Crown research institute ESR, with Forensic Science South Australia.”
. Wakefield involved the issue of whether “Cybergenetics True Allele Casework,” otherwise referred to as TrueAllele, met the Frye standard. That court concluded that it did.
. The Frye hearing consists of two volumes, but each hearing date is consecutively paginated. The January 11 hearing references will be preceded by (I), the January 14 hearing references by (II), and the January 21 hearing references by (III).
. Different computer software has been used by the Lab over the years to calculate a probability ratio, expressed in a mathematical term, e.g., one in 500,000 individuals. In a likelihood ratio, the results are expressed as follows: a match between the suspect and the evidence is (x number) of times more probable than a coincidental match.
. According to the draft guidelines for the Validation of Probabilistic Genotyping Systems, published by SWGDAM and introduced into evidence as People’s exhibit 3,
“A probabilistic genotyping system is comprised of software, or software and hardware, with analytical and statistical function that entail complex formulae and algorithms. Particularly useful for low-level DNA samples and complex mixtures, probabilistic genotyping approaches can reduce subjectivity in the analysis of DNA typing results, as compared to historical methods of mixture interpretation (e.g., deconvolution of the mixture into individual components), and quantifies uncertainty in the analysis.”
. Presumably, defendant is referring to the admonition in People v Leone (
. Having said that, this court is not making a finding on whether he is qualified as an expert to testify at a jury trial about questions related to DNA in general.
. For example, in the only case of which the court is aware that STRmix was the subject of an admissibility hearing, there was testimony from Dr. Simich and Dr. Buckleton, the New Zealand scientist who helped develop STRmix. Some of the other cases reviewed by the court discussed the testimony of several experts, on both sides of the issue, and thus those courts had the benefit of hearing from a greater portion of the relevant scientific community.
. In one of defendant’s many submissions to the court, dated November 3, 2015, there is a letter from Dr. Mark W. Perlin, the chief scientific and executive officer of TrueAllele, to Jerry D. Varnell, Contract Specialist,
. The court in Collins was presented with an impressive group of scientists on both sides of the issue, some of whom testified that the manner in which the drop-in and drop-out rates are assessed at each locus by the FST program is not generally accepted within the DNA community. The court here was presented with no such evidence to the contrary.
. In the court’s research to put this quotation in its proper context, it found the quotation to be, “If your experiment needs statistics, you ought to have done a better experiment.” Without impugning the integrity and brilliance of Rutherford, a Nobel Prize winner in chemistry (although his true field was physics), the court makes two points. First, Rutherford died in 1937 and it is unlikely that he could have envisioned the scope and breadth of scientific advances that have been made in the late twentieth and early twenty-first centuries, particularly in the fields of genetics and computing. Second, Rutherford has also been widely quoted for the following: “The energy produced by the breaking down of the atom is a very poor kind of thing. Anyone who expects a source of power from the transformation of these atoms is talking moonshine.” With the advent of supercomputers, the future is already a thing of the past.