LAJAYVIAN D. DANIELS v. STATE OF FLORIDA

DISTRICT COURT OF APPEAL OF THE STATE OF FLORIDA
FOURTH DISTRICT
LAJAYVIAN D. DANIELS,
Appellant,
v.
STATE OF FLORIDA,
Appellee.
No. 4D19-822
[February 24, 2021]
Appeal from the Circuit Court for the Fifteenth Judicial Circuit, Palm
Beach County; Joseph George Marx, Judge; L.T. Case No. 50-2015-CF-
009320-AXXX-MB.
Carey Haughwout, Public Defender, and Christine C. Geraghty,
Assistant Public Defender, West Palm Beach, for appellant.
Ashley Moody, Attorney General, Tallahassee, and Jessenia J.
Concepcion, Assistant Attorney General, West Palm Beach, for appellee.
CONNER, J.
Appellant, Lajayvian D. Daniels, appeals his convictions and sentences
for first degree murder with a firearm and robbery with a firearm, raising
five issues on appeal. We affirm without discussion the trial court’s rulings
on four issues, but we write to explain our affirmance on the fifth issue
because it concerns a matter of first impression in Florida. The issue
involves the admissibility of expert evidence of a probabilistic genotype
software program used to analyze DNA samples collected while
investigating a crime that contain mixtures of genetic material from
multiple people. We determine the trial court properly admitted the
evidence.
Background
Appellant was indicted for first degree murder with a firearm and
robbery with a firearm in connection with a robbery and shooting resulting
in the death of a gas station employee. For purposes of the issue we
address on appeal, we focus on the pertinent facts and procedural
background regarding the DNA evidence Appellant argues was erroneously
admitted.
The forensic quality assurance manager for the Palm Beach County
Sheriff’s Office (“PBSO”) crime lab testified at trial about DNA samples
collected from the articles discovered during the investigation.
Additionally, she produced DNA reference profiles for four people:
Appellant, the victim, a second suspect, and Appellant’s girlfriend. She
compared those profiles to five DNA samples obtained from five clothing
items believed to be related to the crime. The five samples contained a
mixture of DNA from different persons. Because PBSO did not have
sufficient statistical calculation tools to analyze samples with DNA
mixtures, the forensic quality assurance manager sent the PBSO data files
for those five samples to Cybergenetics, a private lab, for further analysis.
Cybergenetics specializes in DNA mixtures. Cybergenetics was the
developer of TrueAllele, a computer software designed to analyze complex
data to determine the individual profiles of genetic material in DNA
mixtures.
A DNA analyst at Cybergenetics testified at trial that the TrueAllele
software can separate the different genetic types present in samples in
order to calculate match statistics or the level of association between crime
evidence and references. TrueAllele is a probabilistic genotyping system
that relies on Bayesian probability modeling and “Markov Chain” and
“Monte Carlo” statistical sampling. She explained that a Cybergenetics
analyst entered the data compiled by the PBSO DNA analyst into the
TrueAllele program and entered how many people were suspected of
contributing to the mixture. TrueAllele then separated the genetic types
present in the five clothing samples into individual profiles and compared
those profiles to a reference sample to calculate a match statistic. The
Cybergenetics DNA analyst that testified at trial was not the same
Cybergenetics analyst that entered the PBSO lab data into the TrueAllele
software, but she was one of the three reviewers of the TrueAllele analysis
required as a standard protocol by Cybergenetics. She was the witness
who authenticated and discussed at trial the TrueAllele analysis report
admitted into evidence.
The TrueAllele analysis found a statistical match between one clothing
item and Appellant that was 872 trillion times more probable than a
coincidental match to an unrelated person. As to a second item, the match
was 77.1 million times more probable than a coincidental match to an
unrelated person. As to a third item, the match was 194 quadrillion times
more probable than a coincidental match to an unrelated person. As to a
2
fourth item, the match was 789 billion times more probable than a
coincidental match from an unrelated person.
The jury returned a verdict of guilty as charged on both counts. The
trial court sentenced Appellant to concurrent life sentences.
Appellant’s Motion to Exclude TrueAllele Evidence
Prior to trial, the defense filed a “Motion to Exclude the Interpretation
of DNA Mixtures by the TrueAllele Software Due to the Failure to Perform
the Required Internal Validation.” The motion argued that this evidence
did not meet the requirements for admissibility under Daubert v. Merrell
Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993) or Frye v. United States,
293 F. 1013 (D.C. Cir. 1923), because the TrueAllele program was not
“internally validated” prior to being used on the data generated at PBSO.
After acknowledging that TrueAllele evidence has never been litigated in
Florida and the issue of the lack of internal validation has not been ruled
upon in any other jurisdictions regarding the TrueAllele software, the
motion focused on an unpublished New York county court decision issued
in People v. Hillary, 1 which addressed an issue of a lack of internal
validation with regards to STRmix, a different probabilistic genotype
software.
The matter proceeded to a lengthy hearing. The State’s witness at the
hearing was the Cybergenetics DNA analyst who testified about the
development of TrueAllele as a validated probabilistic genotyping system.
She explained the science behind the software and how it works to
determine probabilities. She testified that, in this case, PBSO sent data
on five different samples collected from clothing for interpretation.
The Cybergenetics DNA analyst also testified that she co-authored two
peer-reviewed publications regarding the TrueAllele method of analyzing
DNA and had participated in twelve additional studies involving validating
computer programs or reviewing different aspects of the computer
interpretation. She had testified in nine other cases prior to the instant
case and each time her testimony concerned DNA interpretation and
TrueAllele. She also stated there have been thirty-five validation studies
regarding TrueAllele, and TrueAllele analysts from Cybergenetics have
testified seventy-six times in court. She said that seven validation peer-
1 A copy of People v. Hillary, Decision & Order on DNA Analysis Admissibility,
Indictment No. 2015-15 (N.Y. St. Lawrence Cty. Ct. Aug. 26, 2016) was attached
to Appellant’s motion.
3
reviewed papers on the TrueAllele case work system established that “the
error rates, the sensitivities, specificity and reproducibility of the system
as well as its accuracy,” reflected that the results it was getting made
sense. She testified that there have been sixteen times in multiple states
that TrueAllele was challenged under Daubert or Frye, and in each case,
the results were ruled admissible.
The Cybergenetics DNA analyst further testified that in 2014, she co-
authored a “validation study,” which she explained refers to a test done on
any new method or system or computer method to determine that it is
working as expected, so that it can be tested on a wide variety of data. She
explained there are two types of validation studies: developmental and
internal, which are done to ensure that the system is reliable and establish
any limits or error rates. Developmental validation refers to tests done by
the manufacturer of any method or system to ensure scientific accuracy.
Internal validation, as it relates to forensic science, refers to validation
performed by labs to follow the FBI quality assurance standards for access
to the Combined DNA Index System (“CODIS”). Notably, the Cybergenetics
DNA analyst testified that in eight of the admissibility challenges against
TrueAllele in prior cases where the TrueAllele evidence was ruled
admissible, there was never any internal validation done on the lab from
which the data came nor was the lack of internal validation on a specific
lab’s data an issue for the reliability of the evidence. She noted that a
challenge similar to that raised by the defense in this case to the internal
validation was never raised in any of the prior legal cases of which she was
aware. She moreover testified she was not aware of any studies or papers
reflecting that, without an internal validation, the TrueAllele results are
not scientific.
Concerning People v. Hillary, the Cybergenetics DNA analyst testified
that when using STRmix, internal validation might be necessary, but the
same is not true for TrueAllele:
A: Right, we do not use STRmix, we use TrueAllele.
Q: All right, and so you don’t – do you have to have internal
validation like this Dr. Buckleton was doing where he was
picking and choosing himself the parameters?
A. No, with TrueAllele you just put in – we put in all of the
data. The computer can learn from that data, figure out the
different parameters that he had to input. There is no
calibration that needs to be done with TrueAllele. It can learn
it all from the data. The mathematical model is sophisticated
4
enough that you can put in all the data above the baseline, it
can solve for all those different parameters. You do tell it a few
things, like how many contributors or how long to run for but
not too much more than that.
(emphasis added). In contrast to STRmix, the Cybergenetics DNA analyst
explained:
[W]ith Hillary, the reason why the internal validation would be
important is because they do need those calibration settings
to be able to properly run the [STRmix] program on data from
a specific lab because it’s dependent on knowing information
about that data ahead of time, whereas TrueAllele is not.
The Cybergenetics DNA analyst reiterated that TrueAllele’s sophisticated
scientific model does not require calibration in order to be scientifically
reliable and that this concept had been scientifically tested many times
and is accepted within the community as scientifically reliable.
She further testified that the Scientific Working Group on DNA Analysis
Methods (“SWGDAM”) issues guidelines and recommendations, rather than
mandates, which are then passed on to the FBI to update their quality
assurance guidelines for public crime labs. The Cybergenetics DNA
analyst testified that the latest SWGDAM recommendation, issued in
2015, is applicable to validation of probabilistic genotyping systems, like
TrueAllele. However, she explained that public crime labs utilizing such
systems themselves are required to follow those quality assurance
standards, part of which is the internal validation, in order to access
CODIS, but those standards do not apply to private laboratories, like
Cybergenetics, which do not have access to CODIS. In other words, the
SWGDAM guidance for internal validation is for public labs running their
own tests to be able to maintain CODIS access and verify their handling
of raw specimens. The Cybergenetics DNA analyst further testified:
[V]alidation is important for any scientific method. There are
thirty-five TrueAllele validation studies. Seven of them have
been published in peer reviewed papers. And they test all
kinds of data, including the types of data that were produced
in this case, the same kits, the same sequencers.
(emphasis added). In this regard, the Cybergenetics DNA analyst
explained that internal validation of the in-house data from PBSO was not
required before being used on case work tested at the PBSO lab where
such has been done on the same types of data and where there is no
5
special setting that Cybergenetics would have to input for any kind of data.
The Cybergenetics DNA analyst also added that the SWGDAM guidelines
concerning probabilistic genotyping systems were issued in June 2015
after all of Cybergenetics’s analysis in this case, and the SWGDAM
guidelines state that they are not to be applied retroactively.
On cross examination, the Cybergenetics DNA analyst clarified and
reiterated that Cybergenetics does not produce actual DNA data profiles,
but rather, only interprets them. She further reiterated that Cybergenetics
does comply with the SWGDAM as to all of the applicable guidelines,
noting that it has done developmental validations and that she herself has
done internal validations on TrueAllele on different kinds of data. The
Cybergenetics DNA analyst confirmed that the TrueAllele operator reviews
the data to determine how many contributors there might be, and if there
is a question, for example, as to whether there are three or four
contributors, the operator can have TrueAllele solve for both scenarios and
provide the information for both scenarios. She also confirmed that the
operator can adjust the degradation option, but stated that changing the
input for either factor does not materially affect the result.
The defense called the forensic quality assurance manager for the PBSO
crime lab as a witness at the motion hearing. She testified that because
the PBSO lab is accredited by the FBI, it must follow the FBI’s standards,
which require internal validation of all DNA methods used. She testified
that PBSO now uses STRmix in their lab, for which PBSO conducted an
internal validation. However, she testified that PBSO does not follow every
SWGDAM guideline if it determines it does not apply to the system it is
using or mixtures it is interpreting. The PBSO forensic manager testified
that the data in this case was generated at PBSO using equipment and
kits PBSO routinely uses and had been internally validated prior to use.
She testified that labs using the same testing kits can use different settings
such as the number of analysis cycles they run which can have differences
in the data generated.
On cross examination, the PBSO forensic manager testified she has not
been trained on either the STRmix or the TrueAllele systems. She testified
that: “One of the things that the STRmix does in our system is it uses our
thresholds, it uses our validation, everything that we have goes into the
system so that it takes all of that into account.” However, she contrasted
this with TrueAllele, stating that her understanding of
how TrueAllele works is that you don’t need any of those types
of data, that it’s self – the program self-teaches when it looks
at the data. It doesn’t use stochastic thresholds and things of
6
that nature, that it takes longer to run and it analyzes and
self-teaches with the data.
The defense also called an adjunct professor who teaches forensic
science and who had previously been the quality assurance manager for a
small unrelated private lab. She testified that SWGDAM guidelines are
regarded as the best practices in the field. She explained that validation
is a scientific principle and that for probabilistic genotyping software,
internal validation is “ground truth testing,” where the operator creates a
mixed sample from known individuals and tests the software to see if it
makes the proper inclusions and exclusions. Using in-house data (in this
case, data derived at PBSO) to create the known mixture is important
because the data fluctuations can affect the analysis. She testified that
part of internal validation at the lab generating the data is determining
what is good data and when results are reliable, so as to learn the system
and know its limitations with the individual crime lab’s data. She testified
that things like baseline and instrumentation can impact the data and
that the limitations of the software program need to be defined for the data
at the laboratory that is developing it. She opined that a Cybergenetics
analyst making subjective adjustments in the number of cycle times the
program performs an analysis and other parameters can impact what is
reported and that a Cybergenetics analyst is not in the best position to
know what is the best representation of that data since he or she has not
studied the generating laboratory or the data that is being produced at
that lab. She testified that based on documents she reviewed, there was
no internal validation done using known test sample data created by PBSO
in connection with the TrueAllele system. Finally, she opined that this
internal validation requirement should be imposed as to TrueAllele just as
it is on STRmix, citing to the Hillary case.
On cross examination, the professor admitted she had not prepared
any type of report of her analysis or opinions in this case. Significantly,
she also admitted she had never been associated with TrueAllele or trained
on how it works. She further admitted she had not published any scientific
journals or conducted research on whether TrueAllele was reliable without
internal validation. Moreover, she explained the basis for her opinion on
TrueAllele was her own “self-study” from what was provided by the public
defender and in videos. She also testified she did not perform a validation
to say whether there was a problem with the data.
At the conclusion of the hearing, defense counsel agreed that the
SWGDAM guidelines were not mandated, but argued they were generally
accepted guidelines and best practices which include internal validation,
which was not done in this case by PBSO.
7
The State maintained that TrueAllele was not a lab running forensic
DNA analysis, but was rather a machine calculating mathematical
problems. It pointed to the published TrueAllele validation papers,
asserting that the scientific community has validated TrueAllele and has
embraced its procedures. The State argued that both the Daubert and
Frye standards were met by the evidence it submitted.
After the hearing, the trial court ruled that the TrueAllele expert
testimony was admissible, finding that the TrueAllele analysis results in
this case met the requirements of Frye.
Appellate Analysis
A trial court’s ruling regarding the admissibility of expert testimony is
reviewed on appeal for abuse of discretion. Kemp v. State, 280 So. 3d 81,
88 (Fla. 4th DCA 2019). “However, ‘that discretion is limited by the rules
of evidence.’” Vitiello v. State, 281 So. 3d 554, 559 (Fla. 5th DCA 2019)
(quoting Michael v. State, 884 So. 2d 83, 84 (Fla. 2d DCA 2004)).
When Appellant filed his motion to exclude the DNA interpretations by
the TrueAllele software program, the law was unclear as to whether the
Florida Legislature’s adoption of the Daubert standard for the admission
of expert evidence was constitutional. Hence, Appellant’s motion sought
exclusion of the evidence under both Frye and Daubert. Approximately a
month before the trial court ruled on the motion, our supreme court issued
its opinion in DeLisle v. Crane Co., 258 So. 3d 1219 (Fla. 2018) declaring
the legislative amendment to the Evidence Code unconstitutional. Id. at
1229. Hence, the trial court issued its ruling applying the Frye standard.
Two months after this appeal was filed, however, the supreme court
adopted the legislature’s amendments to the Evidence Code as rules of
procedure, thus changing the evidentiary standard in Florida from Frye to
Daubert. See In re Amendments to Fla. Evidence Code, 278 So. 3d 551,
551–52 (Fla. 2019). Similar to the situation we confronted in Larocca v.
State, 289 So. 3d 492 (Fla. 4th DCA 2020), we apply Daubert to the
resolution of this case because the amendment to the Evidence Code
implementing Daubert is procedural, making its application binding for
our decision. See In re Amendments to Fla. Evidence Code, 278 So. 3d at
552; Larocca, 289 So. 3d at 493 (explaining that “[u]nder Florida’s ‘pipeline
rule,’ the ‘disposition of a case on appeal should be made in accord with
the law in effect at the time of the appellate court’s decision rather than
the law in effect at the time the judgment appealed was
rendered’”(alteration in original) (quoting Kemp, 280 So. 3d at 88)).
8
The Daubert standard is codified under section 90.702, Florida
Statutes, which provides:
If scientific, technical, or other specialized knowledge will
assist the trier of fact in understanding the evidence or in
determining a fact in issue, a witness qualified as an expert
by knowledge, skill, experience, training, or education may
testify about it in the form of an opinion or otherwise, if:
(1) The testimony is based upon sufficient facts or data;
(2) The testimony is the product of reliable principles and
methods; and
(3) The witness has applied the principles and methods
reliably to the facts of the case.
§ 90.702, Fla. Stat. (2018).
Under Daubert, a trial judge is to function as a gatekeeper to “ensure
that any and all scientific testimony or evidence admitted is not only
relevant, but reliable.” 509 U.S. at 589 (emphasis added). The gatekeeping
function is “‘to ensure that speculative, unreliable expert testimony does
not reach the jury’ under the mantle of reliability that accompanies the
appellation ‘expert testimony.’” Kemp, 280 So. 3d at 88 (quoting Rink v.
Cheminova, Inc., 400 F.3d 1286, 1291 (11th Cir. 2005)). Although there
is no definitive list of factors for the court to consider in making this
determination, the Daubert court laid out several observations it deemed
appropriate for consideration of the reliability inquiry, including: (1)
“whether [the] theory or technique . . . can be (and has been) tested”; (2)
“whether the theory or technique has been subjected to peer review and
publication”; (3) “in the case of a particular scientific technique, the court
ordinarily should consider the known or potential rate of error”; and (4)
“general acceptance.” Daubert, 509 U.S. at 593–94.
The crux of Appellant’s argument on appeal is that the failure to
internally validate the TrueAllele software using a test sample of PBSO-
generated DNA data prior to using the program for case work rendered the
TrueAllele analysis results unreliable under Daubert, and therefore, the
trial court abused its discretion in permitting such evidence. Appellant
supports his position by contending that there was no dispute that the
SWGDAM guidelines (which require internal validation) are the best
practices, and that failure to use the acknowledged best practices means
the results obtained from the TrueAllele software would not be generally
9
accepted in the DNA scientific community and, therefore, the TrueAllele
analysis in this case was unreliable. Additionally, Appellant asserts that
the internal validation study done with regard to one lab does not relieve
the best practice of performing internal validations of other labs. Thus,
Appellant asserts that Cybergenetics’s internal validation studies on DNA
samples for other labs does not alleviate the failure to perform an internal
validation study as to the PBSO lab.
Appellant acknowledges there are no appellate cases from this state or
any other jurisdictions determining that the failure of a lab to perform an
internal validation study renders the results of a TrueAllele analysis
inadmissible. Instead, Appellant relies on the unpublished New York
county court ruling regarding the inadmissibility of an analysis using the
STRmix software because the law enforcement agency did not internally
validate the software. In seeking reversal, Appellant argues that while the
Cybergenetics DNA analyst testified TrueAllele did not require calibration
like STRmix, the operator still has to decide how many different people
contributed to the DNA mixture and whether to turn off the degradation
feature, such that these subjective factors inputted by the operator are the
reason internal validation is necessary. However, we conclude that
Appellant’s argument fails to establish the trial court abused its discretion
in permitting the evidence.
Upon review of the transcript of the hearing on the motion to exclude
and, in particular, the pertinent background facts identified above, we are
satisfied that the trial court properly assessed and concluded that the DNA
statistical interpretation performed by the TrueAllele software program
was reliable after considering: (1) the theory or technique has been tested;
(2) the theory or technique has been subjected to peer review and
publication; (3) the known or potential rate of error for the program; and
(4) the general acceptance of the program. See Daubert, 509 U.S. at 593–
94. We are also satisfied that the trial court gave specific consideration to
Appellant’s argument regarding the lack of internal validation but
concluded the argument and evidence did not merit excluding the
TrueAllele evidence. Although the trial court applied the Frye standard,
we conclude the trial court’s analysis and conclusions would have been
the same under the Daubert standard. 2 In terms of assessing reliability
2The Frye standard is that “expert testimony should be deduced from generally
accepted scientific principles.” DeLisle, 258 So. 3d at 1225. As discussed above,
one of the indicia of reliability under Daubert is “general acceptance” in the
scientific community. We note that in denying the motion to exclude, the trial
10
under either standard, the focal point of Appellant’s argument in the trial
court and on appeal has been the reliability factor of “general acceptance.”
It is particularly significant that Appellant has cited no appellate decision
in Florida or elsewhere to support his argument. It is also particularly
significant that the defense expert in this case was not sufficiently familiar
with the TrueAllele software to effectively opine as to how the failure to
internally validate the software using PBSO-generated test data
compromised the reliability of the analysis of the DNA samples collected
from clothing during the criminal investigation of this case. The trial court
did not abuse its discretion or violate any evidentiary rule by determining
it was not convinced that the lack of internal validation in this case made
the results unreliable. We therefore affirm on this issue, as well as the
other issues raised on appeal.
Affirmed.
WARNER and FORST, JJ., concur.
* * *
Not final until disposition of timely filed motion for rehearing.
court references and discusses evidence regarding all of the Daubert reliability
factors.
11