Single-cell investigative genetics: Single-cell data produces genotype distributions concentrated at the true genotype across all mixture complexities.

In the absence of a suspect the forensic aim is investigative, and the focus is one of discerning what genotypes best explain the evidence. In traditional systems, the list of candidate genotypes may become vast if the sample contains DNA from many donors or the information from a minor contributor is swamped by that of major contributors, leading to lower evidential value for a true donor's contribution and, as a result, possibly overlooked or inefficient investigative leads. Recent developments in single-cell analysis offer a way forward, by producing data capable of discriminating genotypes. This is accomplished by first clustering single-cell data by similarity without reference to a known genotype. With good clustering it is reasonable to assume that the scEPGs in a cluster are of a single contributor. With that assumption we determine the probability of a cluster's content given each possible genotype at each locus, which is then used to determine the posterior probability mass distribution for all genotypes by application of Bayes' rule. A decision criterion is then applied such that the sum of the ranked probabilities of all genotypes falling in the set is at least 1-α. This is the credible genotype set and is used to inform database search criteria. Within this work we demonstrate the salience of single-cell analysis by performance testing a set of 630 previously constructed admixtures containing up to 5 donors of balanced and unbalanced contributions. We use scEPGs that were generated by isolating single cells, employing a direct-to-PCR extraction treatment, amplifying STRs that are compliant with existing national databases and applying post-PCR treatments that elicit a detection limit of one DNA copy. We determined that, for these test data, 99.3% of the true genotypes are included in the 99.8% credible set, regardless of the number of donors that comprised the mixture. We also determined that the most probable genotype was the true genotype for 97% of the loci when the number of cells in a cluster was at least two. Since efficient investigative leads will be borne by posterior mass distributions that are narrow and concentrated at the true genotype, we report that, for this test set, 47,900 (86%) loci returned only one credible genotype and of these 47,551 (99%) were the true genotype. When determining the LR for true contributors, 91% of the clusters rendered LR>1018, showing the potential of single-cell data to positively affect investigative reporting.

The analogy between DNA kinship and DNA mixture evaluation, with applications for the interpretation of likelihood ratios produced by possibly imperfect models.

Two main applications of forensic DNA analysis are the investigation of possible relatedness and the investigation whether a person left DNA in a trace. Both of these are usually carried out by the calculation of likelihood ratios. In the kinship case, it is standard to let the likelihood ratio express the support in favour of the investigated relatedness versus no relatedness, and in the investigation of traces, one by default compares the hypothesis that the person of interest contributed DNA, versus that he is unrelated to any of the actual contributors. In both cases however, we can also view the probabilistic procedure as an inference of the profile of the person we look for: in other words, in both cases we carry out probabilistic genotyping. In this article we use this general analogy to develop various more specific analogies between kinship and mixture likelihood ratios. These analogies help to understand the concepts that play a role, and also to understand the importance of the statistical modeling needed for DNA mixtures. In this article, we apply our findings to consider what we can and cannot conclude from a likelihood ratio in favour of contribution to a mixed DNA profile, if that is computed by a model whose specifics are not entirely known to us, or where we do not know whether they provide a good description of the stochastic effects involved in the generation of DNA trace profiles. We show that, if unrelated individuals are adequately modeled, we can give bounds on how often LR's coming from certain types of black box models may arise, both for persons who are actual contributors and who are unrelated. In particular we show that no model, provided it satisfies basic requirements, can overestimate the evidence found for actual contributors both often and strongly.

Evaluation of glass evidence at activity level: A new distribution for the background population.

For evidence evaluation of the physicochemical properties of glass at activity level a well-known formula introduced by Evett & Buckleton [1,2] is commonly used. Parameters in this formula are, amongst others, the probability in a background population to find on somebody's clothing the observed number of glass sources and the probability in a background population to find on somebody's clothing a group of fragments with the same size as the observed matching group. Recently, for efficiency reasons, the Netherlands Forensic Institute changed its methodology to measure not all the glass fragments but a subset of glass fragments found on clothing. Due to the measurement of subsets, it is difficult to get accurate estimates for these parameters in this formula. We offer a solution to this problem. The heart of the solution consists of relaxing the assumption of conditional independence of group sizes of background fragments, and modelling the probability of an allocation of background fragments into groups given a total number of background fragments by a two-parameter Chinese restaurant process (CRP) [3]. Under the assumption of random sampling of fragments to be measured from recovered fragments in the laboratory, parameter values for the Chinese restaurant process may be estimated from a relatively small dataset of glass in other relevant cases. We demonstrate this for a dataset of glass fragments collected from upper garments in casework, show model fit and provide a prototypical calculation of an LR at activity level accompanied with a parameter sensitivity analysis for reasonable ranges of the CRP parameter values. Considering that other laboratories may want to measure subsets as well, we believe this is an important alternative approach to the evaluation of numerical LRs for glass analyses at activity level.

A top-down approach to DNA mixtures.

Presently, there exist many different models and algorithms for determining, in the form of a likelihood ratio, whether there is evidence that a person of interest contributed to a mixed trace profile. These methods have in common that they model the whole trace, hence all its contributors, which leads to the computation time being mostly determined by the number of contributors that is assumed. At some point, these calculations are no longer feasible. We present another approach, in which we target the contributors of the mixture in the order of their contribution. With this approach the calculation time now depends on how many contributors are queried. This means that any trace can be subjected to calculations of likelihood ratios in favor of being a relatively prominent contributor, and we can choose not to query it for all its contributors, e.g., if that is computationally not feasible, or not relevant for the case. We do so without using a quantitative peak height model, i.e., we do not define a peak height distribution. Instead, we work with subprofiles derived from the full trace profile, carrying out likelihood ratio calculations on these with a discrete method. This lack of modeling makes our method widely applicable. The results with our top-down method are slightly conservative with respect to the one of a continuous model, and more so as we query less and less prominent contributors. We present results on mixtures with known contributors and on research data, analyzing traces with plausibly 6 or more contributors. If a top-k of most prominent contributors is targeted, it is not necessary to know how many other contributors there are for LR calculations, and the more prominent the queried contributor is relatively to all others, the less the evidential value depends on the specifics of a chosen peak height model. For these contributors the qualitative statement that more input DNA leads to larger peaks suffices. The evidential value for a comparison with minor contributors on the other hand, potentially depends much more on the chosen model. We also conclude that a trace's complexity, as meaning its (in)ability to yield large LR's that are not too model-dependent, is not measured by its number of contributors; rather, it is the equality of contribution that makes it harder to obtain strong evidence.

DNAxs/DNAStatistX: Development and validation of a software suite for the data management and probabilistic interpretation of DNA profiles.

The data management, interpretation and comparison of sets of DNA profiles can be complex, time-consuming and error-prone when performed manually. This, combined with the growing numbers of genetic markers in forensic identification systems calls for expert systems that can automatically compare genotyping results within (large) sets of DNA profiles and assist in profile interpretation. To that aim, we developed a user-friendly software program or DNA eXpert System that is denoted DNAxs. This software includes features to view, infer and match autosomal short tandem repeat profiles with connectivity to up and downstream software programs. Furthermore, DNAxs has imbedded the 'DNAStatistX' module, a statistical library that contains a probabilistic algorithm to calculate likelihood ratios (LRs). This algorithm is largely based on the source code of the quantitative probabilistic genotyping system EuroForMix [1]. The statistical library, DNAStatistX, supports parallel computing which can be delegated to a computer cluster and enables automated queuing of requested LR calculations. DNAStatistX is written in Java and is accessible separately or via DNAxs. Using true and non-contributors to DNA profiles with up to four contributors, the DNAStatistX accuracy and precision were assessed by comparing the DNAStatistX results to those of EuroForMix. Results were the same up to rare differences that could be attributed to the different optimizers used in both software programs. Implementation of dye specific detection thresholds resulted in larger likelihood values and thus a better explanation of the data used in this study. Furthermore, processing time, robustness of DNAStatistX results and the circumstances under which model validations failed were examined. Finally, guidelines for application of the software are shared as an example. The DNAxs software is future-proof as it applies a modular approach by which novel functionalities can be incorporated.

Response paper to "The likelihood of encapsulating all uncertainty": The relevance of additional information for the LR.

In this response paper, part of the Virtual Special Issue on "Measuring and Reporting the Precision of Forensic Likelihood Ratios", we further develop our position on likelihood ratios which we described previously in Berger et al. (2016) "The LR does not exist". Our exposition is inspired by an example given in Martire et al. (2016) "On the likelihood of encapsulating all uncertainty", where the consequences of obtaining additional information on the LR were discussed. In their example, two experts use the same data in a different way, and the LRs of these experts change differently when new data are taken into account. Using this example as a starting point we will demonstrate that the probability distribution for the frequency of the characteristic observed in trace and reference material can be used to predict how much an LR will change when new data become available. This distribution can thus be useful for such a sensitivity analysis, and address the question of whether to obtain additional data or not. But it does not change the answer to the original question of how to update one's prior odds based on the evidence, and it does not represent an uncertainty on the likelihood ratio based on the current data.

Accurate assessment of the weight of evidence for DNA mixtures by integrating the likelihood ratio.

Several methods exist for weight of evidence calculations on DNA mixtures. Especially if dropout is a possibility, it may be difficult to estimate mixture specific parameters needed for the evaluation. For semi-continuous models, the LR for a person to have contributed to a mixture depends on the specified number of contributors and the probability of dropout for each. We show here that, for the semi-continuous model that we consider, the weight of evidence can be accurately obtained by applying the standard statistical technique of integrating the likelihood ratio against the parameter likelihoods obtained from the mixture data. This method takes into account all likelihood ratios belonging to every choice of parameters, but LR's belonging to parameters that provide a better explanation to the mixture data put in more weight into the final result. We therefore avoid having to estimate the number of contributors or their probabilities of dropout, and let the whole evaluation depend on the mixture data and the allele frequencies, which is a practical advantage as well as a gain in objectivity. Using simulated mixtures, we compare the LR obtained in this way with the best informed LR, i.e., the LR using the parameters that were used to generate the data, and show that results obtained by integration of the LR approximate closely these ideal values. We investigate both contributors and non-contributors for mixtures with various numbers of contributors. For contributors we always obtain a result close to the best informed LR whereas non-contributors are excluded more strongly if a smaller dropout probability is imposed for them. The results therefore naturally lead us to reconsider what we mean by a contributor, or by the number of contributors.

The LR does not exist.

More than 40years ago, De Finetti warned that probability is a misleading misconception when regarded as objectively existing exterior to the mind. According to De Finetti, probabilities are necessarily subjective, and quantify our belief in the truth of events in the real world. Given evidence of a shared feature of a trace and an accused, we apply this framework to assign an evidential value to this correspondence. Dividing 1 by the objectively existing proportion of the population sharing that feature would give that evidential value - expressed as a likelihood ratio (LR) - only if that proportion were known. As in practice the proportion can only be estimated, this leads some to project their sampling uncertainty - or precision - associated with the estimated proportion onto the likelihood ratio, and to report an interval. Limited data should limit our LR however, because as we will demonstrate the LR is given by what we know about the proportion rather than by the unknown proportion itself. Encapsulating all uncertainty - including sampling uncertainty of the proportion - our LR reflects how much information we have retrieved from the feature regarding the trace's origin, based on our present knowledge. Not an interval but a number represents this amount of information, equal to the logarithm of the LR. As long as we know how to interpret the evidence with a well-defined probabilistic model, we know what our evidence is worth.

The likelihood ratio as a random variable for linked markers in kinship analysis.

The likelihood ratio is the fundamental quantity that summarizes the evidence in forensic cases. Therefore, it is important to understand the theoretical properties of this statistic. This paper is the last in a series of three, and the first to study linked markers. We show that for all non-inbred pairwise kinship comparisons, the expected likelihood ratio in favor of a type of relatedness depends on the allele frequencies only via the number of alleles, also for linked markers, and also if the true relationship is another one than is tested for by the likelihood ratio. Exact expressions for the expectation and variance are derived for all these cases. Furthermore, we show that the expected likelihood ratio is a non-increasing function if the recombination rate increases between 0 and 0.5 when the actual relationship is the one investigated by the LR. Besides being of theoretical interest, exact expressions such as obtained here can be used for software validation as they allow to verify the correctness up to arbitrary precision. The paper also presents results and advice of practical importance. For example, we argue that the logarithm of the likelihood ratio behaves in a fundamentally different way than the likelihood ratio itself in terms of expectation and variance, in agreement with its interpretation as weight of evidence. Equipped with the results presented and freely available software, one may check calculations and software and also do power calculations.

Exclusion probabilities and likelihood ratios with applications to mixtures.

The statistical evidence obtained from mixed DNA profiles can be summarised in several ways in forensic casework including the likelihood ratio (LR) and the Random Man Not Excluded (RMNE) probability. The literature has seen a discussion of the advantages and disadvantages of likelihood ratios and exclusion probabilities, and part of our aim is to bring some clarification to this debate. In a previous paper, we proved that there is a general mathematical relationship between these statistics: RMNE can be expressed as a certain average of the LR, implying that the expected value of the LR, when applied to an actual contributor to the mixture, is at least equal to the inverse of the RMNE. While the mentioned paper presented applications for kinship problems, the current paper demonstrates the relevance for mixture cases, and for this purpose, we prove some new general properties. We also demonstrate how to use the distribution of the likelihood ratio for donors of a mixture, to obtain estimates for exceedance probabilities of the LR for non-donors, of which the RMNE is a special case corresponding to L R>0. In order to derive these results, we need to view the likelihood ratio as a random variable. In this paper, we describe how such a randomization can be achieved. The RMNE is usually invoked only for mixtures without dropout. In mixtures, artefacts like dropout and drop-in are commonly encountered and we address this situation too, illustrating our results with a basic but widely implemented model, a so-called binary model. The precise definitions, modelling and interpretation of the required concepts of dropout and drop-in are not entirely obvious, and we attempt to clarify them here in a general likelihood framework for a binary model.

p-values should not be used for evaluating the strength of DNA evidence.

Recently, p-values have been suggested to explain the strength of a likelihood ratio that evaluates DNA evidence. It has been argued that likelihood ratios would be difficult to explain in court and that p-values would offer an alternative that is easily explained. In this article, we argue that p-values should not be used in this context. p-Values do not directly relate to the strength of the evidence. The likelihood ratio measures the strength of the evidence, while the p-value measures how rare it is to find evidence that is equally strong or stronger, which is something fundamentally different. In addition, a p-value is not always unambiguous. To illustrate our arguments, we present several examples from forensic genetics.

Optimal strategies for familial searching.

Familial searching is the process of finding potential relatives of the donor of a crime scene profile in a DNA database. Several authors have proposed strategies for generating candidate lists of potential relatives. This paper reviews four strategies and investigates theoretical properties as well as empirical behavior, using a comprehensive simulation study on mock databases. The effectiveness of a familial search is shown to highly depend on the case profile as well as on the tuning parameters. We give recommendations for proceeding in an optimal way and on how to choose tuning parameters both in general and on a case-by-case basis. Additionally we treat searching heterogeneous databases (not all profiles comprise the same loci) and composite searching for multiple types of kinship. An R-package for reproducing results in a particular case is released to help decision-making in familial searching.

Exclusion probabilities and likelihood ratios with applications to kinship problems.

In forensic genetics, DNA profiles are compared in order to make inferences, paternity cases being a standard example. The statistical evidence can be summarized and reported in several ways. For example, in a paternity case, the likelihood ratio (LR) and the probability of not excluding a random man as father (RMNE) are two common summary statistics. There has been a long debate on the merits of the two statistics, also in the context of DNA mixture interpretation, and no general consensus has been reached. In this paper, we show that the RMNE is a certain weighted average of inverse likelihood ratios. This is true in any forensic context. We show that the likelihood ratio in favor of the correct hypothesis is, in expectation, bigger than the reciprocal of the RMNE probability. However, with the exception of pathological cases, it is also possible to obtain smaller likelihood ratios. We illustrate this result for paternity cases. Moreover, some theoretical properties of the likelihood ratio for a large class of general pairwise kinship cases, including expected value and variance, are derived. The practical implications of the findings are discussed and exemplified.