Validation of a DNA mixture statistics tool incorporating allelic drop-out and drop-in

DNA mixture analysis is a current topic of discussion in the forensics literature. Of particular interest is how to approach mixtures where allelic drop-out and/or drop-in may have occurred. The Office of Chief Medical Examiner (OCME) of The City of New York has developed and validated the Forensic Statistical Tool (FST), a software tool for likelihood ratio analysis of forensic DNA samples, allowing for allelic drop-out and drop-in. FST can be used for single source samples and for mixtures of DNA from two or three contributors, with or without known contributors. Drop-out and drop-in probabilities were estimated empirically through analysis of over 2000 amplifications of more than 700 mixtures and single source samples. Drop-out rates used by FST are a function of the Identifiler^® locus, the quantity of template DNA amplified, the number of amplification cycles, the number of contributors to the sample, and the approximate mixture ratio (either unequal or approximately equal). Drop-out rates were estimated separately for heterozygous and homozygous genotypes. Drop-in rates used by FST are a function of number of amplification cycles only.FST was validated using 454 mock evidence samples generated from DNA mixtures and from items handled by one to four persons. For each sample, likelihood ratios (LRs) were computed for each true contributor and for each profile in a database of over 1200 non-contributors. A wide range of LRs for true contributors was obtained, as true contributors’ alleles may be labeled at some or all of the tested loci. However, the LRs were consistent with OCME's qualitative assessments of the results. The second set of data was used to evaluate FST LR results when the test sample in the prosecution hypothesis of the LR is not a contributor to the mixture. With this validation, we demonstrate that LRs generated using FST are consistent with, but more informative than, OCME's qualitative sample assessments and that LRs for non-contributors are appropriately assigned.     

ProbRank: An efficient DNA database search method for complex mixtures per a quantitative likelihood ratio model

Searching a DNA Database with a DNA profile from an evidentiary trace can provide investigative leads in a forensic case. Various searching approaches exist such as conventional methods based on matching alleles or more advanced methods computing likelihood ratios (LR) while considering drop-in and drop-out. Here we examine the potential of using a quantitative LR model (EuroForMix model incorporated in ProbRank method) that takes peak heights into account in comparison to a qualitative LR model (LRmix model implemented in SmartRank method). Both methods present DNA database candidates in order of decreasing LR. Especially regarding minor contributors in complex mixtures, the method using the quantitative model outperforms the method using the qualitative model in terms of sensitivity and specificity as more true donors and less adventitious matches are retrieved. ProbRank is to be implemented in DNAStatistX and is sufficiently fast for daily use.     

DNA commission of the International Society of Forensic Genetics: Recommendations on the evaluation of STR typing results that may include drop-out and/or drop-in using probabilistic methods

DNA profiling of biological material from scenes of crimes is often complicated because the amount of DNA is limited and the quality of the DNA may be compromised. Furthermore, the sensitivity of STR typing kits has been continuously improved to detect low level DNA traces. This may lead to (1) partial DNA profiles and (2) detection of additional alleles. There are two key phenomena to consider: allelic or locus ‘drop-out’, i.e. ‘missing’ alleles at one or more genetic loci, while ‘drop-in’ may explain alleles in the DNA profile that are additional to the assumed main contributor(s). The drop-in phenomenon is restricted to 1 or 2 alleles per profile. If multiple alleles are observed at more than two loci then these are considered as alleles from an extra contributor and analysis can proceed as a mixture of two or more contributors. Here, we give recommendations on how to estimate probabilities considering drop-out, Pr(D), and drop-in, Pr(C). For reasons of clarity, we have deliberately restricted the current recommendations considering drop-out and/or drop-in at only one locus. Furthermore, we offer recommendations on how to use Pr(D) and Pr(C) with the likelihood ratio principles that are generally recommended by the International Society of Forensic Genetics (ISFG) as measure of the weight of the evidence in forensic genetics. Examples of calculations are included. An Excel spreadsheet is provided so that scientists and laboratories may explore the models and input their own data.     

Interpreting low template DNA profiles

We discuss the interpretation of DNA profiles obtained from low template DNA samples. The most important challenge to interpretation in this setting arises when either or both of “drop-out” and “drop-in” create discordances between the crime scene DNA profile and the DNA profile expected under the prosecution allegation. Stutter and unbalanced peak heights are also problematic, in addition to the effects of masking from the profile of a known contributor. We outline a framework for assessing such evidence, based on likelihood ratios that involve drop-out and drop-in probabilities, and apply it to two casework examples. Our framework extends previous work, including new approaches to modelling homozygote drop-out and uncertainty in allele calls for stutter, masking and near-threshold peaks. We show that some current approaches to interpretation, such as ignoring a discrepant locus or reporting a “Random Man Not Excluded” (RMNE) probability, can be systematically unfair to defendants, sometimes extremely so. We also show that the LR can depend strongly on the assumed value for the drop-out probability, and there is typically no approximation that is useful for all values. We illustrate that ignoring the possibility of drop-in is usually unfair to defendants, and argue that under circumstances in which the prosecution relies on drop-out, it may be unsatisfactory to ignore any possibility of drop-in.     

Exploratory data analysis for the interpretation of low template DNA mixtures

The interpretation of DNA mixtures has proven to be a complex problem in forensic genetics. In particular, low template DNA samples, where alleles can be missing (allele drop-out), or where alleles unrelated to the crime-sample are amplified (allele drop-in), cannot be analysed with classical approaches such as random man not excluded or random match probability. Drop-out, drop-in, stutters and other PCR-related stochastic effects, create uncertainty about the composition of the crime-sample, making it difficult to attach a weight of evidence when (a) reference sample(s) is (are) compared to the crime-sample. In this paper, we use a probabilistic model to calculate likelihood ratios when there is uncertainty about the composition of the crime-sample. This model is essentially exploratory in the sense that it allows the exploration of LRs when two key-parameters, drop-out and drop-in are varied within their plausible ranges of variation. We build on the work of Curran et al. [8], and improve their probabilistic model to allow more flexibility in the way the model parameters are applied. Two new main modifications are brought to their model: (i) different drop-out probabilities can be applied to different contributors, and (ii) different parameters can be used under the prosecution and the defence hypotheses. We illustrate how the LRs can be explored when the drop-out and drop-in parameters are varied, and suggest the use of Monte Carlo simulations to derive plausible ranges for the probability of drop-out. Although the model is suited for both high and low template samples, we illustrate the advantages of the exploratory approach through two DNA mixtures (involving two and at least three individuals) with low template components.     

Separation of SNP profiles from DNA mixtures with two contributors via massively parallel sequencing technology

ABSTRACT

Massively parallel sequencing technology offers the opportunity to analyse forensically challenging samples, such as degraded samples and mixtures. In the current study, we developed a perl-based pipeline to separate the DNA mixture into its components and to predict the most probable single nucleotide polymorphism (SNP) genotypes of each contributor to the mixed profile. We examined the usefulness of this method by detecting both artificially constructed DNA mixtures and mixtures from crime cases using the Precision ID Identity Panel on the Ion PGM platform. The separated genotypes of mixtures were validated by genotypes of each of the donors detected independently. The results indicate that the method performed well in identifications of both the artificially constructed mixtures and case-type mixtures, even when the two contributors are immediate relatives (mother and son), which demonstrated the practical usefulness of this method in forensic casework. Our research presents an efficient and different strategy for identification and paternity testing of DNA mixtures in forensic genetics.     

Internal validation of STRmix™ for the interpretation of single source and mixed DNA profiles

The interpretation of DNA evidence can entail analysis of challenging STR typing results. Genotypes inferred from low quality or quantity specimens, or mixed DNA samples originating from multiple contributors, can result in weak or inconclusive match probabilities when a binary interpretation method and necessary thresholds (such as a stochastic threshold) are employed. Probabilistic genotyping approaches, such as fully continuous methods that incorporate empirically determined biological parameter models, enable usage of more of the profile information and reduce subjectivity in interpretation. As a result, software-based probabilistic analyses tend to produce more consistent and more informative results regarding potential contributors to DNA evidence. Studies to assess and internally validate the probabilistic genotyping software STRmix™ for casework usage at the Federal Bureau of Investigation Laboratory were conducted using lab-specific parameters and more than 300 single-source and mixed contributor profiles. Simulated forensic specimens, including constructed mixtures that included DNA from two to five donors across a broad range of template amounts and contributor proportions, were used to examine the sensitivity and specificity of the system via more than 60,000 tests comparing hundreds of known contributors and non-contributors to the specimens. Conditioned analyses, concurrent interpretation of amplification replicates, and application of an incorrect contributor number were also performed to further investigate software performance and probe the limitations of the system. In addition, the results from manual and probabilistic interpretation of both prepared and evidentiary mixtures were compared.The findings support that STRmix™ is sufficiently robust for implementation in forensic laboratories, offering numerous advantages over historical methods of DNA profile analysis and greater statistical power for the estimation of evidentiary weight, and can be used reliably in human identification testing. With few exceptions, likelihood ratio results reflected intuitively correct estimates of the weight of the genotype possibilities and known contributor genotypes. This comprehensive evaluation provides a model in accordance with SWGDAM recommendations for internal validation of a probabilistic genotyping system for DNA evidence interpretation     

Regression models for DNA-mixtures

This paper deals with the statistical interpretation of DNA mixture evidence. The conventional methods used in forensic casework today use something like 16 STR-markers. Power can be increased by rather using SNP-markers. New statistical methods are then needed, and we present a regression framework. The basic idea is that the traditional forensic hypotheses, commonly denoted H_D and H_P, are replaced by parametric versions: a person contributes to a mixture if and only if the fraction he contributes is greater than 0. This contributed fraction is a parameter of the regression model. The regression model uses the peak heights directly and there is no need to specify or estimate the number of contributors to the mixture. Also, drop-in and drop-out pose no principal problems.Data from 25 controlled blinded experiments were used to test the model. The number of contributors varied between 2 and 5, and the fractions contributed ranged from 0.01 to 0.99. The fractions were accurately estimated by the regression analyses. There were no false positives (i.e., in no cases were non-contributors declared to contributors). Some false negatives occurred for fractions of 0.1 or lower. Simulations were performed to test the model further. The analyses show that useful estimates can be obtained from a relatively small number of SNP-markers. Reasonable results are achieved using 300 markers which is close to the 313 SNPs in the controlled experiment. Increasing the number of SNPs, the analyses demonstrate that individuals contributing as little as 1% can reliably be detected, which suggests that cases beyond the reach of conventional forensic methods today can be reported.     

Inference about the number of contributors to a DNA mixture: Comparative analyses of a Bayesian network approach and the maximum allele count method

In the forensic examination of DNA mixtures, the question of how to set the total number of contributors (N) presents a topic of ongoing interest. Part of the discussion gravitates around issues of bias, in particular when assessments of the number of contributors are not made prior to considering the genotypic configuration of potential donors. Further complication may stem from the observation that, in some cases, there may be numbers of contributors that are incompatible with the set of alleles seen in the profile of a mixed crime stain, given the genotype of a potential contributor. In such situations, procedures that take a single and fixed number contributors as their output can lead to inferential impasses. Assessing the number of contributors within a probabilistic framework can help avoiding such complication. Using elements of decision theory, this paper analyses two strategies for inference on the number of contributors. One procedure is deterministic and focuses on the minimum number of contributors required to ‘explain’ an observed set of alleles. The other procedure is probabilistic using Bayes’ theorem and provides a probability distribution for a set of numbers of contributors, based on the set of observed alleles as well as their respective rates of occurrence. The discussion concentrates on mixed stains of varying quality (i.e., different numbers of loci for which genotyping information is available). A so-called qualitative interpretation is pursued since quantitative information such as peak area and height data are not taken into account. The competing procedures are compared using a standard scoring rule that penalizes the degree of divergence between a given agreed value for N, that is the number of contributors, and the actual value taken by N. Using only modest assumptions and a discussion with reference to a casework example, this paper reports on analyses using simulation techniques and graphical models (i.e., Bayesian networks) to point out that setting the number of contributors to a mixed crime stain in probabilistic terms is, for the conditions assumed in this study, preferable to a decision policy that uses categoric assumptions about N.     

Validation and assessment of the Investigator® 24plex QS kit for forensic casework application: Comparison with the PowerPlex® fusion system and GlobalFiler™ PCR amplification kits

Casework evidence samples are likely to be placed under diverse and harsh environments as compared to quantified DNA samples including serial-diluted standard DNA samples. Internal validation of a novel STR kit using casework evidence sample, which is conducted according to various conditions such as DNA contamination and degradation, is crucial before being used as a forensic application. Therefore, this study aimed to elucidate the reliability of the Investigator® 24plex QS kit through DNA derived from casework evidence and to assess whether it is applicable to STR analysis together with PowerPlex® Fusion System and GlobalFiler™ PCR Amplification Kit. DNA was extracted from 189 casework evidence samples in a total of 77 cases. The mismatch of the allelic size of this kit through allelic sizing precision test, was suitable according to ENFSI guidelines. All heterozygous balance of the three kits were above 0.6 recommended value of ENFSI guideline. The number of allele drop-in was most frequent in the GlobalFiler™ PCR Amplification Kit. In addition, the number of allele drop-out was most frequent in the Investigator® 24plex QS kit. The cutoff concentration of DNA detected in three kits of one complete STR was approximately 45 pg/μL on average. Despite of several limitations, the Investigator® 24plex QS kit is considered to have the capability to be used for STR analysis of casework evidence samples.     

On the exact distribution of the numbers of alleles in DNA mixtures

When more than one individual contributes biological material to a forensic stain, the resulting DNA type is termed a DNA mixture. DNA mixtures occur frequently in forensic genetic casework, and in recent years, much research has been devoted to this subject. This paper presents a derivation of the exact distribution of the number of alleles for any number of profiles and investigated loci. The per locus number of observed alleles is of interest as it indicates the plausible range on the number of contributors. Hence, by specifying a prior distribution on the number of contributors, the locus distribution may be used to assess the number of contributors. Furthermore, the total number of alleles across all loci is used by some forensic geneticists to estimate the probability that an allele has failed to be detected (allelic drop-out).     

Interpreting Y chromosome STR haplotype mixture

Mixture interpretation is a challenging problem in forensic DNA analyses. The interpretation of Y short tandem repeat (STR) haplotype mixtures, due to a lack of recombination, differs somewhat from that of the autosomal DNA markers and is more complex. We describe approaches for calculating the probability of exclusion (PE) and likelihood ratio (LR) methods to interpret Y-STR mixture evidence with population substructure incorporated. For a mixture sample, first, all possible contributor haplotypes in a reference database are listed as a candidate list. The PE is the complement of the summation of the frequencies of haplotypes in the candidate list. The LR method compares the probabilities of the evidence given alternative hypotheses. The hypotheses are possible explanations for the mixture. Population substructure may be further incorporated in likelihood calculation. The maximum number of contributors is based on the candidate list and the computing complexity is polynomial. Additionally, mixtures were simulated by combining two or three 16 Y-STR marker haplotypes derived from the US forensic Y-STR database. The average PE was related to the size of database. With a database comprised of 500 haplotypes an average PE value of at least 0.995 can be obtained for two-person mixtures. The PE decreases with an increasing number of contributors to the mixture. Using the total sample population, the average number of candidate haplotypes of two-person mixtures is 3.73 and 95% mixtures have less than or equal to 10 candidate haplotypes. More than 98.7% of two-person mixtures can only be explained by the haplotype combinations that mixtures are composed. These values are generally higher for three-person mixtures. A small proportion of three-person mixture can also be explained by only two haplotypes.     

An assessment of the performance of the probabilistic genotyping software EuroForMix: Trends in likelihood ratios and analysis of Type I & II errors

Continuous probabilistic genotyping software enables the interpretation of highly complex DNA profiles that are prone to stochastic effects and/or consist of multiple contributions. The process of introducing probabilistic genotyping into an accredited forensic laboratory requires testing, validation, documentation and training. Documents that include guidelines and/or requirements have been published in order to guide forensic laboratories through this extensive process and there has been encouragements to share the results obtained from internal laboratory studies. To this end, we present the results obtained from the quantitative probabilistic genotyping system EuroForMix applied to mixed DNA profiles with known contributions mixed in known proportions, levels of allele sharing and levels of allelic drop-out. The mixtures were profiled using the PowerPlex® Fusion 6C (PPF6C) kit. Using these mixtures, 427 Hp-true tests and 408 Hd-true tests were performed. In the Hd-true tests, non-contributors were selected deliberately to a have large overlap with the alleles within the mixture and worst-case scenarios were examined in which a simulated relative of one of the true donors was considered as the person of interest under the prosecution hypothesis. The effects of selecting different EuroForMix modelling options, the use of PCR replicates, allelic drop-out, and varying the assigned number of contributors were examined. Instances of Type I and Type II errors are discussed. In addition 330 likelihood ratio results from EuroForMix are compared to the semi-continuous model LRmix Studio. Results demonstrate the performance and trends of EuroForMix when applied to PPF6C profiles.     

Open source software EuroForMix can be used to analyse complex SNP mixtures

A series of two- and three-person mixtures of varying dilutions were prepared and analysed with Life Technologies’ HID-Ion AmpliSeq™ Identity Panel v2.2 using the Ion PGM™ massively parallel sequencing (MPS) system. From this panel we used 134 autosomal SNPs. Using the reference samples of three donors, we evaluated the strength of evidence with likelihood ratio (LR) calculations using the open-source quantitative EuroForMix program and compared the results with a previous study using a qualitative software (LRmix). SNP analysis is a special case of STRs, restricted to a maximum of two alleles per locus. We showed that simple two-person mixtures can be readily analysed with both LRmix and Euroformix, but the performance of three- or more person mixtures is generally inefficient with LRmix. Taking account of the “peak height” information, by substituting ‘sequence read’ coverage values from the MPS data for each SNP allele, greatly improves the discrimination between true and non-contributors. The higher the mixture proportion (Mx) of the person of interest is, the higher the LR. Simulation experiments (up to six contributors) showed that the strength of the evidence is dependent upon Mx, but relatively insensitive to the number of contributors. If a higher number of loci were multiplexed, the analysis of mixtures would be much improved, because the extra information would enable lower Mx values to be evaluated. In summary, incorporating the 'sequence read' (coverage) into the quantitative model shows a significant benefit over the qualitative approach. Calculations are quite fast (six seconds for three contributors).     

Facile semi-automated forensic body fluid identification by multiplex solution hybridization of NanoString® barcode probes to specific mRNA targets

A DNA profile from the perpetrator does not reveal, per se, the circumstances by which it was transferred. Body fluid identification by mRNA profiling may allow extraction of contextual ‘activity level’ information from forensic samples. Here we describe the development of a prototype multiplex digital gene expression (DGE) method for forensic body fluid/tissue identification based upon solution hybridization of color-coded NanoString^® probes to 23 mRNA targets. The method identifies peripheral blood, semen, saliva, vaginal secretions, menstrual blood and skin. We showed that a simple 5 min room temperature cellular lysis protocol gave equivalent results to standard RNA isolation from the same source material, greatly enhancing the ease-of-use of this method in forensic sample processing.We first describe a model for gene expression in a sample from a single body fluid and then extend that model to mixtures of body fluids. We then describe calculation of maximum likelihood estimates (MLEs) of body fluid quantities in a sample, and we describe the use of likelihood ratios to test for the presence of each body fluid in a sample. Known single source samples of blood, semen, vaginal secretions, menstrual blood and skin all demonstrated the expected tissue-specific gene expression for at least two of the chosen biomarkers. Saliva samples were more problematic, with their previously identified characteristic genes exhibiting poor specificity. Nonetheless the most specific saliva biomarker, HTN3, was expressed at a higher level in saliva than in any of the other tissues.Crucially, our algorithm produced zero false positives across this study's 89 unique samples. As a preliminary indication of the ability of the method to discern admixtures of body fluids, five mixtures were prepared. The identities of the component fluids were evident from the gene expression profiles of four of the five mixtures. Further optimization of the biomarker ‘CodeSet’ will be required before it can be used in casework, particularly with respect to increasing the signal-to-noise ratio of the saliva biomarkers. With suitable modifications, this simplified protocol with minimal hands on requirement should facilitate routine use of mRNA profiling in casework laboratories.     

Quantification of forensic genetic evidence: Comparison of results obtained by qualitative and quantitative software for real casework samples

To overcome the multifactorial complexity associated with the analysis and interpretation of the capillary electrophoresis results of forensic mixture samples, probabilistic genotyping methods have been developed and implemented as software, based on either qualitative or quantitative models. The former considers the electropherograms’ qualitative information (detected alleles), whilst the latter also takes into account the associated quantitative information (height of allele peaks). Both models then quantify the genetic evidence through the computation of a likelihood ratio (LR), comparing the probabilities of the observations given two alternative and mutually exclusive hypotheses.In this study, the results obtained through the qualitative software LRmix Studio (v.2.1.3), and the quantitative ones: STRmix™ (v.2.7) and EuroForMix (v.3.4.0), were compared considering real casework samples. A set of 156 irreversibly anonymized sample pairs (GeneMapper files), obtained under the scope of former cases of the Portuguese Scientific Police Laboratory, Judiciary Police (LPC-PJ), were independently analyzed using each software. Sample pairs were composed by (i) a mixture profile with either two or three estimated contributors, and (ii) a single contributor profile associated. In most cases, information on 21 short tandem repeat (STR) autosomal markers were considered, and the majority of the single-source samples could not be a priori excluded as belonging to a contributor to the paired mixture sample. This inter-software analysis shows the differences between the probative values obtained through different qualitative and quantitative tools, for the same input samples. LR values computed in this work by quantitative tools showed to be generally higher than those obtained by the qualitative. Although the differences between the LR values computed by both quantitative software showed to be much smaller, STRmix™ generated LRs are generally higher than those from EuroForMix. As expected, mixtures with three estimated contributors showed generally lower LR values than those obtained for mixtures with two estimated contributors.Different software products are based on different approaches and mathematical or statistical models, which necessarily result in the computation of different LR values. The understanding by the forensic experts of the models and their differences among available software is therefore crucial. The better the expert understands the methodology, the better he/she will be able to support and/or explain the results in court or any other area of scrutiny.     

Interpreting a major component from a mixed DNA profile with an unknown number of minor contributors

Modern interpretation strategies typically require an assignment of the number of contributors (N) to a DNA profile. This can prove to be a difficult task, particularly when dealing with higher order mixtures or mixtures where one or more contributors have donated low amounts of DNA. Differences in the assigned N at interpretation can lead to differences in the likelihood ration (LR). If the number of contributors cannot reasonably be assigned, then an interpretation of the profile may not be able to be progressed.In this study, we investigate mixed DNA profiles of varying complexity and interpret them altering the assigned N. We assign LRs for true- and non- contributors and compare the results given different assignments of N over a range of mixture proportions. When a component of a mixture had a proportion of at least 10%, a ratio of at least 1.5:1 to the next highest component, and a DNA amount (as determined by STRmix™) of at least 50 rfu, the LR of the component for a true contributor was not significantly affected by varying N and was therefore suitable for interpretation and the assignment of an LR. LRs produced for minor contributors were found to vary significantly as the assigned N was changed. These heuristics may be used to identify profiles suitable for interpretation.     

Distinguishing between donors and their relatives in complex DNA mixtures with binary models

While likelihood ratio calculations were until the recent past limited to the evaluation of mixtures in which all alleles of all donors are present in the DNA mixture profile, more recent methods are able to deal with allelic dropout and drop-in. This opens up the possibility to obtain likelihood ratios for mixtures where this was not previously possible, but it also means that a full match between the alleged contributor and the crime stain is no longer necessary. We investigate in this article what the consequences are for relatives of the actual donors, because they typically share more alleles with the true donor than an unrelated individual. We do this with a semi-continuous binary approach, where the likelihood ratios are based on the observed alleles and the dropout probabilities for each donor, but not on the peak heights themselves. These models are widespread in the forensic community. Since in many cases a simple model is used where a uniform dropout probability is assumed for all (or for all unknown) contributors, we explore the extent to which this alters the false positive probabilities for relatives of donors, compared to what would have been obtained with the correct probabilities of dropout for each donor.     

Automated estimation of the number of contributors in autosomal short tandem repeat profiles using a machine learning approach

The number of contributors (NOC) to (complex) autosomal STR profiles cannot be determined with absolute certainty due to complicating factors such as allele sharing and allelic drop-out. The precision of NOC estimations can be improved by increasing the number of (highly polymorphic) markers, the use of massively parallel sequencing instead of capillary electrophoresis, and/or using more profile information than only the allele counts.In this study, we focussed on machine learning approaches in order to make maximum use of the profile information. To this end, a set of 590 PowerPlex® Fusion 6C profiles with one up to five contributors were generated from a total of 1174 different donors. This set varied for the template amount of DNA, mixture proportion, levels of allele sharing, allelic drop-out and degradation. The dataset contained labels with known NOC and was split into a training, test and hold-out set. The training set was used to optimize ten different algorithms with selection of profile characteristics. Per profile, over 250 characteristics, denoted ‘features’, were calculated. These features were based on allele counts, peak heights and allele frequencies. The features that were most related to the NOC were selected based on partial correlation using the training set. Next, the performance of each model (=combination of features plus algorithm) was examined using the test set. A random forest classifier with 19 features, denoted the ‘RFC19-model’ showed best performance and was selected for further validation. Results showed improved accuracy compared to the conventional maximum allele count approach and an in-house nC-tool based on the total allele count. The method is extremely fast and regarded useful for application in forensic casework.     

Maximizing allele detection: Effects of analytical threshold and DNA levels on rates of allele and locus drop-out

Interpretation of DNA evidence depends upon the ability of the analyst to accurately compare the DNA profile obtained from an item of evidence and the DNA profile of a standard. This interpretation becomes progressively more difficult as the number of ‘drop-out’ and ‘drop-in’ events increase. Analytical thresholds (AT) are typically selected to ensure the false detection of noise is minimized. However, there exists a tradeoff between the erroneous labeling of noise as alleles and the false non-detection of alleles (i.e. drop-out). In this study, the effect ATs had on both types of error was characterized. Various ATs were tested, where three relied upon the analysis of baseline signals obtained from 31 negative samples. The fourth AT was determined by utilizing the relationship between RFU signal and DNA input. The other ATs were the commonly employed 50, 150 and 200 RFU thresholds. Receiver Operating Characteristic (ROC) plots showed that although high ATs completely negated the false labeling of noise, DNA analyzed with ATs derived using analysis of the baseline signal exhibited the lowest rates of drop-out and the lowest total error rates. In another experiment, the effect small changes in ATs had on drop-out was examined. This study showed that as the AT increased from ～10 to 60 RFU, the number of heterozygous loci exhibiting the loss of one allele increased. Between ATs of 60 and 150 RFU, the frequency of allelic drop-out remained constant at 0.27 (±0.02) and began to decrease when ATs of 150 RFU or greater were utilized. In contrast, the frequency of heterozygous loci exhibiting the loss of both alleles consistently increased with AT. In summary, for samples amplified with less than 0.5 ng of DNA, ATs derived from baseline analysis of negatives were shown to decrease the frequency of drop-out by a factor of 100 without significantly increasing rates of erroneous noise detection.