Comparison of the performance of different models for the interpretation of low level mixed DNA profiles.

DNA analyses from forensic casework samples commonly result in complex DNA profiles. Often, these profiles consist of multiple contributors and display multiple stochastic events such as peak height imbalance, allelic or locus drop-out, allelic drop-in, and excessive or indistinguishable stutter. This increased complexity has established a need for more sophisticated methods of DNA mixture interpretation. This study compares the effectiveness of statistical models in the interpretation of artificially created low template two person mixed DNA profiles at varying proportions and template quantities. Two binary models (combined probability of inclusion and random match probability), a semicontinuous (Lab retriever), and continuous model (STRmix™) were compared. Generally, as the sophistication of the models increases, the power of discrimination increases. Differences in discrimination often correlate to each model's ability to use observed data effectively. Binary models require static thresholds resulting in unused data and outliers that may lead to difficult or incorrect interpretation. Semicontinuous and continuous models eliminate the stochastic threshold, however Lab Retriever does not account for stochastic events beyond drop-out and drop-in leading to possible less effective use of the data. STRmix™ incorporates all stochastic events listed above into the calculation making the most effective use of the observed data.

Modeling forward stutter: toward increased objectivity in forensic DNA interpretation.

Forward stutter, or over stutter, one repeat unit length larger than the parent allele (N + 1 stutter), is a relatively rare product of the PCR amplification of STRs used in forensic DNA analysis. We have investigated possible explanatory variables for the occurrence and size of forward stutter for four different autosomal multiplexes. In addition, we have investigated models used to predict the expected heights of forward stutter. For all tetra and penta-nucleotide repeats we can find no correlation between allelic peak height, marker, or longest uninterrupted sequence in the allele. The data fit a gamma distribution with no explanatory variables. For the single trinucleotide repeat present in two of the four multiplexes (D22S1045) forward stutter is much more common and the best explanatory variable appears to be back stutter height. This suggests some fundamental cocausation of high backward and forward stutter for this locus.

The 'factor of two' issue in mixed DNA profiles.

A commonly used idea in forensic fields is known as the 'hierarchy of propositions'. DNA analysts commonly report at the sub-source level in the hierarchy. This means that they simply comment on the probability of the evidence for the given propositions that consider contributors that lead to a DNA profile and not on the source of specific biological components, not the activity that led to the transfer or the offence that is reported to have occurred. However DNA analysts also commonly report at a level even lower than the sub-source level. In this 'sub-sub-source' level only reference comparisons to components of a mixture are reported. The difference between the sub-source level and sub-sub-source level is the difference between comparing an individual to a mixture as a whole, or comparing them to only one component of a mixture. This idea has been expressed in the past as the 'two trace' problem or the 'factor of two' problem. With the advent of expert systems that can provide a measure of weight of evidence in the form of a likelihood ratio (LR) for any mixture, resolvable or not, the distinction between these two levels becomes more important. In this paper we explore how the LR can be constructed to report correctly at the sub-source level, by taking contributor orders and genotype set orders into account. We include worked examples of the LR calculation to help explain this confusing issue.

A comparison of statistical models for the analysis of complex forensic DNA profiles.

Complex mixtures and LtDNA profiles are difficult to interpret. As yet there is no consensus within the forensic biology community as to how these profiles should be interpreted. This paper is a review of some of the current interpretation models, highlighting their weaknesses and strengths. It also discusses what a forensic biologist requires in an interpretation model and if this can be realistically executed under current justice systems.

Helping formulate propositions in forensic DNA analysis.

The Bayesian paradigm is the preferred approach to evidence interpretation. It requires the evaluation of the probability of the evidence under at least two propositions. The value of the findings (i.e., our LR) will depend on these propositions and the case information, so it is crucial to identify which propositions are useful for the case at hand. Previously, a number of principles have been advanced and largely accepted for the evaluation of evidence. In the evaluation of traces involving DNA mixtures there may be more than two propositions possible. We apply these principles to some exemplar situations. We also show that in some cases, when there are no clear propositions or no defendant, a forensic scientist may be able to generate explanations to account for observations. In that case, the scientist plays a role of investigator, rather than evaluator. We believe that it is helpful for the scientist to distinguish those two roles.

Extending the discussion on inconsistency in forensic decisions and results.

The subject of inter- and intra-laboratory inconsistency was recently raised in a commentary by Itiel Dror. We re-visit an inter-laboratory trial, with which some of the authors of this current discussion were associated, to diagnose the causes of any differences in the likelihood ratios (LRs) assigned using probabilistic genotyping software. Some of the variation was due to different decisions that would be made on a case-by-case basis, some due to laboratory policy and would hence differ between laboratories, and the final and smallest part was the run-to-run difference caused by the Monte Carlo aspect of the software used. However, the net variation in LRs was considerable. We believe that most laboratories will self-diagnose the cause of their difference from the majority answer and in some, but not all instances will take corrective action. An inter-laboratory exercise consisting of raw data files for relatively straightforward mixtures, such as two mixtures of three or four persons, would allow laboratories to calibrate their procedures and findings.

Sequencing-induced artefacts in NGS STR data.

Significant progress has been made in recent years in the development of techniques for Next Generation Sequencing (NGS), or Massively Parallel Sequencing (MPS), of forensically relevant short tandem repeat (STR) loci. However, as these technologies are investigated and adopted by forensic laboratories, new challenges unfold that require further scrutiny. In the analysis of DNA profiles generated using the MiSeq FGx sequencing system, we have observed noise sequences with relatively high readcounts that are challenging to distinguish from genuine alleles. These high read count noise sequences appear as allele sequences with one or a few substituted bases compared to a known allele sequence within the profile. An examination of ForenSeq DNA Signature Prep Kit STR noise sequences revealed that the substituted base of a parent allele can align to the same position on the sequence across noise sequences. This suggests that these substitution events occur at specific positions within the amplicon, resulting in multiple noise reads with substitutions at the same position. Mapping of the noise events onto the original raw read positions revealed a high number of events, or "noise spikes", occurring at specific positions within a given sequencing run. These noise spikes affected reads across the entire run, agnostic of locus or sample, while the position, occurrence, and amplitude of the spikes differed across runs. The majority of noise sequences with high read counts in a DNA profile were generated from base changes at these spike positions, and could be classified as "noise spike artefacts". In this paper we present evidence of the noise spike artefacts and their genesis during the sequencing process in the sequencing-by-synthesis (SBS) cycles, as well as the methods developed to detect them. The information and methods will assist laboratories with detecting noise spikes in MiSeq FGx sequencing runs, differentiating authentic allele sequences from noise spike artefacts, and developing protocols for analyst review and handling of MiSeq FGx data.

Estimation of population specific values of theta for sequence-based STR profiles.

We describe the estimation of θ (theta) values from autosomal STR sequencing data for five metapopulations. The data were compiled from 20 publications and included 39 datasets comprising a total of 7005 samples. The estimates are suitable for use within the calculation of match probabilities in forensic casework. We also have constructed a phylogenetic tree using this data that aligns with our understanding of human evolution.

A collaborative study on the precision of the Markov chain Monte Carlo algorithms used for DNA profile interpretation.

Several fully continuous probabilistic genotyping software (PGS) use Markov chain Monte Carlo algorithms (MCMC) to assign weights to different proposed genotype combinations at a locus. Replicate interpretations of the same profile in these software are expected not to produce identical weights and likelihood ratio (LR) values due to the Monte Carlo aspect. This paper reports a detailed precision study under reproducibility conditions conducted as a collaborative exercise across the National Institute of Standards and Technology (NIST), Federal Bureau of Investigation (FBI), and Institute of Environmental Science and Research (ESR). Replicate interpretations generated across the three laboratories used the same input files, software version, and settings but different random number seed and different computers. This work demonstrates that using different computers to analyze replicate interpretations does not contribute to any variations in LR values. The study quantifies the magnitude of differences in the assigned LRs that is only due to run-to-run MCMC variability and addresses the potential explanations for the observed differences.

A diagnosis of the primary difference between EuroForMix and STRmix™.

There is interest in comparing the output, principally the likelihood ratio, from the two probabilistic genotyping software EuroForMix (EFM) and STRmix™. Many of these comparison studies are descriptive and make little or no effort to diagnose the cause of difference. There are fundamental differences between EFM and STRmix™ that are causative of the largest set of likelihood ratio differences. This set of differences is for false donors where there are many instances of LRs just above or below 1 for EFM that give much lower LRs in STRmix™. This is caused by the separate estimation of parameters such as allele height variance and mixture proportion using MLE under Hp and Ha for EFM. This can result in very different estimations of these parameters under Hp and Ha . It results in a departure from calibration for EFM in the region of LRs just above and below 1.

A comparison of likelihood ratios with and without assuming relatedness for DNA mixtures interpreted using a continuous model.

When evaluating support for the contribution of a person of interest (POI) to a mixed DNA sample, it is generally assumed that the mixture contributors are unrelated to the POI and to each other. In practice, there may be situations where this assumption is violated, for instance if two mixture contributors are siblings. The effect on the likelihood ratio of (in)correctly assuming relatedness between mixture contributors has previously been investigated using simulation studies based on simplified models ignoring peak heights. We revisit this problem using a simulation study that applies peak height models both in the simulation and mixture interpretation part of the study. Specifically, we sample sets of mixtures comprising both related and unrelated contributors and evaluate support for the contribution of the mixture donors as well as unrelated persons with and without incorporating an assumption of relatedness. The results show, consistent with earlier studies, that including a correct assumption of relatedness increases the capacity of the probabilistic genotyping system to distinguish between mixture donors and unrelated persons. Any effect of the relatedness is found to depend strongly on the mixture ratio. We further show that the results do not change materially when a sub-population correction is applied. Finally, we suggest and discuss a likelihood ratio approach that considers relatedness between mixture contributors using a prior probability.

An Investigation into Compound Likelihood Ratios for Forensic DNA Mixtures.

Simple propositions are defined as those with one POI and the remaining contributors unknown under Hp and all unknown contributors under Ha. Conditional propositions are defined as those with one POI, one or more assumed contributors, and the remaining contributors (if any) unknown under Hp, and the assumed contributor(s) and N unknown contributors under Ha. In this study, compound propositions are those with multiple POI and the remaining contributors unknown under Hp and all unknown contributors under Ha. We study the performance of these three proposition sets on thirty-two samples (two laboratories × four NOCs × four mixtures) consisting of four mixtures, each with N = 2, N = 3, N = 4, and N = 5 contributors using the probabilistic genotyping software, STRmix™. In this study, it was found that conditional propositions have a much higher ability to differentiate true from false donors than simple propositions. Compound propositions can misstate the weight of evidence given the propositions strongly in either direction.

Developmental validation of STRmix™ NGS, a probabilistic genotyping tool for the interpretation of autosomal STRs from forensic profiles generated using NGS.

We describe the developmental validation of the probabilistic genotyping software - STRmix™ NGS - developed for the interpretation of forensic DNA profiles containing autosomal STRs generated using next generation sequencing (NGS) also known as massively parallel sequencing (MPS) technologies. Developmental validation was carried out in accordance with the Scientific Working Group on DNA Analysis Methods (SWGDAM) Guidelines for the Validation of Probabilistic Genotyping Systems and the International Society for Forensic Genetics (ISFG) recommendations and included sensitivity and specificity testing, accuracy, precision, and the interpretation of case-types samples. The results of developmental validation demonstrate the appropriateness of the software for the interpretation of profiles developed using NGS technology.

A tool for simulating single source and mixed DNA profiles.

Simulation studies play an important role in the study of probabilistic genotyping systems, as a low cost and fast alternative to in vitro studies. With ongoing calls for further study of the behaviour of probabilistic genotyping systems, there is a continuous need for such studies. In most cases, researchers use simplified models, for example ignoring complexities such as peak height variability due to lack of availability of advanced tools. We fill this void and describe a tool that can simulate DNA profiles in silico for the validation and investigation of probabilistic genotyping software. Contributor genotypes are simulated by randomly sampling alleles from selected allele frequencies. Some or all contributors may be related to a pedigree and the genotypes of non-founders are obtained by random gene dropping. The number of contributors per profile, and ranges for parameters such as DNA template amount and degradation parameters can be configured. Peak height variability is modelled using a lognormal distribution or a gamma distribution. Profile behaviour of simulated profiles is shown to be broadly similar to laboratory generated profiles though the latter shows more variation. Simulation studies do not remove the need for experimental data. The tool has been made available as an R-package named simDNAmixtures.

Investigation into the effect of mixtures comprising related people on non-donor likelihood ratios, and potential practises to mitigate providing misleading opinions.

The interpretation of mixtures containing related individuals can be difficult due to allele sharing between the contributors. Challenges include the assignment of the number of contributors (NoC) to the mixture with the under assignment of NoC resulting in false exclusions of true donors. Non-donating relatives of the true contributors to mixtures of close relatives can result in likelihood ratios supporting their adventitious inclusion within the mixture. We examine the effect of non-donor likelihood ratios on mixtures of first order relatives. Mixtures of full siblings and parent-child were created by mixing the DNA from known family members in vitro, or by in silico simulation. Mixtures were interpreted using the probabilistic genotyping software STRmix™ and likelihood ratios were assigned for the true donors and non-donors who were either further relatives of the true donors or unrelated to the true donors. The two donor balanced mixtures deconvoluted straightforwardly when analysed as NoC = 2 giving approximately the experimental design 1:1 ratio. When analysed as NoC = 3 a very large number of non-donor genotypes produced LRs close to 1 including many instances of adventitious support. The in vitro three donor balanced mixtures proved difficult to assign as NoC = 3 by a blind examination of the profile. It is likely that many of these would be misassigned as NoC = 2. The analysis of the in vitro and in silico mixtures assuming NoC = 3 with no use of a conditioning profile or with the use of a conditioning profile but without informed priors on the mixture proportions (Mx priors) was ineffective. If the profile can be assigned as NoC = 3 then assignment of the Mx priors is straightforward. This analysis gave no false exclusions. Adventitious support did happen for relatives with high allele sharing. Adventitious support was not observed for any unrelated non-donors. The analysis of the three-person mixtures as NoC = 2 produced many false exclusions and fewer instances of adventitious support. The three donor unbalanced mixtures could all be assigned as NoC= 3. Analysis without Mx priors produced an alternate genotype explanation.

Evaluating DNA Mixtures with Contributors from Different Populations Using Probabilistic Genotyping.

It is common practice to evaluate DNA profiling evidence with likelihood ratios using allele frequency estimates from a relevant population. When multiple populations may be relevant, a choice has to be made. For two-person mixtures without dropout, it has been reported that conservative estimates can be obtained by using the Person of Interest's population with a θ value of 3%. More accurate estimates can be obtained by explicitly modelling different populations. One option is to present a minimum likelihood ratio across populations; another is to present a stratified likelihood ratio that incorporates a weighted average of likelihoods across multiple populations. For high template single source profiles, any difference between the methods is immaterial as far as conclusions are concerned. We revisit this issue in the context of potentially low-level and mixed samples where the contributors may originate from different populations and study likelihood ratio behaviour. We first present a method for evaluating DNA profiling evidence using probabilistic genotyping when the contributors may originate from different ethnic groups. In this method, likelihoods are weighted across a prior distribution that assigns sample donors to ethnic groups. The prior distribution can be constrained such that all sample donors are from the same ethnic group, or all permutations can be considered. A simulation study is used to determine the effect of either assumption on the likelihood ratio. The likelihood ratios are also compared to the minimum likelihood ratio across populations. We demonstrate that the common practise of taking a minimum likelihood ratio across populations is not always conservative when FST=0. Population stratification methods may also be non-conservative in some cases. When FST>0 is used in the likelihood ratio calculations, as is recommended, all compared approaches become conservative on average to varying degrees.

Exploring likelihood ratios assigned for siblings of the true mixture contributor as an alternate contributor.

Relatives tend to have more DNA in common than unrelated people. The closer the biological relationship, the higher the chance of alleles being identical by descent between the individuals. Therefore, when considering a mixed DNA profile, close relatives of the true contributor may not always be excluded as a possible contributor to a mixture due to allele sharing. In these situations, it might be more appropriate under the alternate proposition to consider that the DNA could have originated from a relative of the person of interest rather than an unrelated individual. The probabilistic genotyping software STRmix™ automatically provides LRs considering close biological relatives as alternate sources of the DNA. In this paper, we investigate the support for siblings of the true contributor to a mixture (who are not present in the mixture themselves). We interpret the mixtures and assign LRs using STRmix™ and investigate whether the resulting LRs could be used to indicate whether the true contributor could be a sibling of the POI. Most siblings will have one or more alleles that are not observed in the mixture profile. Support for siblings to have contributed can only occur when allelic dropout is a possibility at the loci where the siblings have alleles that are not observed in the profile. In these data, that was only observed in components with assigned template of 588 rfu or less.

A mixed DNA profile controversy revisited.

Semaan et al. (J Forensic Res, 2020, 11, 453) discuss a mock case "where eight different individuals [P1 through P8 ] could not be excluded in a mixed DNA analysis. Even though … expert DNA mixture analysis software was used." Two of these are the true donors. The LRs reported are incorrect due to the incorrect entry of propositions into LRmix Studio. This forced the software to account for most of the alleles as drop-in, resulting in LRs 60-70 orders of magnitude larger than expected. P1 , P2 , P4 , P5 , and P8 can be manually excluded using peak heights. This has relevance when using LRmix which does not use peak heights. We extend the work using the same two reference genotypes who were the true contributors as Semaan et al. (J Forensic Res, 2020, 11, 453). We simulate three two-donor mixtures with peak heights using these two genotypes and analyze using STRmix™. For the simulated 1:1 mixture, one of the non-donors' LRs supported him being a contributor when no conditioning was used. When considered in combination with any other potential donors (i.e., with conditioning), this non-donor was correctly eliminated. For the 3:1 mixture, all results correctly supported that the non-donors were not contributors. The low-template 4:1 mixture LRs with no conditioning showed support for all eight profiles as donors. However, the results from pair-wise conditioning showed that only the two ground truth donors had LRs supporting that they were contributors to the mixture. We recommend the use of peak heights and conditioning profiles, as this allows better sensitivity and specificity even when the persons share many alleles.

Evaluating DNA evidence possibly involving multiple (mixed) samples, common donors and related contributors.

Forensic DNA profiling is used in various circumstances to evaluate support for two competing propositions with the assignment of a likelihood ratio. Many software implementations exist that tackle a range of inference problems spanning identification and relationship testing. We propose a flexible likelihood ratio framework that caters to inference problems in forensic genetics. The framework allows for investigation of the degree of support for the contribution of multiple persons to multiple samples allowing for persons to be related according to a pedigree, including inbred relationships. We explain how a number of routine as well as more complex problems can be treated within this framework.

Probabilistic interpretation of the Amelogenin locus.

We describe a method to assign weights to genotype combinations at the Amelogenin locus. It is a typical practise in forensic laboratories that once the weight exceeds a threshold (such as 99 %), then they can be considered to be resolved enough to interpret (for example to load onto a database). We found that unless an individual is a clear major (or minor) contributor, the genotype weights do not typically exceed 99 % for any genotype. LRs have not been traditionally assigned for the Amelogenin locus and are small compared to an LR assigned for a modern day STR multiplex where, for a more discriminatory locus, the per locus LR for a resolved contributor could be in the order of tens or hundreds. The method described uses per contributor template values for a previously interpreted profile. The discrimination power is restricted, with a maximum possible LR of 2 for a fully resolved genotype, due to the limited number of alleles and hence genotypes and assuming equal proportions of genders in the population. However, it has a good power to exclude when the component is well resolved and non-concordant with a POI.