Probabilistic expert systems for forensic inference from DNA markers in horses: applications to confirm genealogies with lack of genetic data.

Microsatellites have been used for parentage testing and individual identification in forensic science because they are highly polymorphic and show abundant sequences dispersed throughout most eukaryotic nuclear genomes. At present, genetic testing based on DNA technology is used for most domesticated animals, including horses, to confirm identity, to determine parentage, and to validate registration certificates. But if genetic data of one of the putative parents are missing, verifying a genealogy could be questionable. The aim of this paper is to illustrate a new approach to analyze complex cases of disputed relationship with microsatellites markers. These cases were solved by analyzing the genotypes of the offspring and other horses' genotypes in the pedigrees of the putative dam/sire with probabilistic expert systems (PESs). PES was especially efficient in supplying reliable, error-free Bayesian probabilities in complex cases with missing pedigree data. One of these systems was developed for forensic purposes (FINEX program) and is particularly valuable in human analyses. We applied this program to parentage analysis in horses, and we will illustrate how different cases have been successfully worked out.

Paternity testing that involves a DNA mixture.

Here we analyse a complex disputed paternity case, where the DNA of the putative father was extracted from his corpse that had been inhumed for over 20 years. This DNA was contaminated and appears to be a mixture of at least two individuals. Furthermore, the mother's DNA was not available. The DNA mixture was analysed so as to predict the most probable genotypes of each contributor. The major contributor's profile was then used to compute the likelihood ratio for paternity. We also show how to take into account a dropout allele and the possibility of mutation in paternity testing.

'Stochastic' effects at balanced mixtures: a calibration study.

DNA mixtures are challenging not only at low template DNA level but also at highly balanced quantitative ratio. In this latter case, interpretation may be complicated by the joint action of combinatorial uncertainty and stochastic effects of the PCR. We explore this particular and so far little noticed aspect of mixture interpretation by first providing a complete quantitative combinatorial analysis of the two-person mixture model (2PM) at highly balanced ratio of contributors, and then by carrying out a calibration study of the 2PM model on good quality experimental mixtures. The calibration tests provided the evidence for the existence of irregular distribution of peak heights, that can misguide the correct genotype assignment at high template ratios too. Repeating the experiment, performing Bayesian analysis to the whole evidence and developing a careful joint prediction of all plausible genotype datasets is highly mandatory in these cases, prior to set evidentiary LRs and use them in court.

Joint Bayesian analysis of forensic mixtures.

Evaluation of series of PCR experiments referring to the same evidence is not infrequent in a forensic casework. This situation is met when 'series of results in mixture' (EPGs produced by reiterating PCR experiments over the same DNA mixture extract) have to be interpreted or when 'potentially related traces' (mixtures that can have contributors in common) require a combined interpretation. In these cases, there can be uncertainty on the genotype assignment, since: (a) more than one genotype combination fall under the same peak profile; (b) PCR preferential amplification alters pre-PCR allelic proportions; (c) other, more unpredictable technical problems (dropouts/dropins, etc.) take place. The uncertainty in the genotype assignment is in most cases addressed by empirical methods (selection of just one particular profile; extraction of consensual or composite profiles) that disregard part of the evidence. Genotype assignment should conversely take advantage from a joint Bayesian analysis (JBA) of all STRs peak areas generated at each experiment. This is the typical case of Bayesian analysis in which adoption of object-oriented Bayesian networks (OOBNs) could be highly helpful. Starting from experimentally designed mixtures, we created typical examples of 'series of results in mixture' of 'potentially related traces'. JBA was some administered to the whole peak area evidence, by specifically tailored OOBNs models, which enabled genotype assignment reflecting all the available evidence. Examples of a residual ambiguity in the genotype assignment came to light at assumed genotypes with partially overlapping alleles (for example: AB+AC→ABC). In the 'series of results in mixture', this uncertainty was in part refractory to the joint evaluation. Ambiguity was conversely dissipated at the 'potentially related' trace example, where the ABC allelic scheme at the first trace was interpreted together with other unambiguous combinations (ABCD; AB) at the related trace. We emphasize the need to carry out extensive, blind sensitivity tests specifically addressing the residual ambiguity that arises from overlapping results mixed at various quantitative ratios.

Using forensic microsatellites to decipher the genetic structure of linguistic and geographic isolates: A survey in the eastern Italian Alps.

The study of geographically and/or linguistically isolated populations could represent a potential area of interaction between population and forensic genetics. These investigations may be useful to evaluate the suitability of loci which have been selected using forensic criteria for bio-anthropological studies. At the same time, they give us an opportunity to evaluate the efficiency of forensic tools for parentage testing in groups with peculiar allele frequency profiles. Within the frame of a long-term project concerning Italian linguistic isolates, we studied 15 microsatellite loci (Identifiler kit) comprising the CODIS panel in 11 populations from the north-eastern Italian Alps (Veneto, Trentino and Friuli Venezia Giulia regions). All our analyses of inter-population differentiation highlight the genetic distinctiveness of most Alpine populations comparing them either to each other or with large and non-isolated Italian populations. Interestingly, we brought to light some aspects of population genetic structure which cannot be detected using unilinear polymorphisms. In fact, the analysis of genotypic disequilibrium between loci detected signals of population substructure when all the individuals of Alpine populations are pooled in a single group. Furthermore, despite the relatively low number of loci analyzed, genetic differentiation among Alpine populations was detected at individual level using a Bayesian method to cluster multilocus genotypes. Among the various populations studied, the four linguistic minorities (Fassa Valley, Luserna, Sappada and Sauris) showed the most pronounced diversity and signatures of a peculiar genetic ancestry. Finally, we show that database replacement may affect estimates of probability of paternity even when the local database is replaced by another based on populations which share a common genetic background but which differ in their demographic history. These findings point to the importance of considering the demographic and cultural profile of populations in forensic applications, even in a context of substantial genetic homogeneity such as that of European populations.

Autopsy investigation and Bayesian approach to coronary artery disease in victims of motor-vehicle accidents.

BACKGROUND: Each year, 1.2 million people die worldwide as a result of motor-vehicle accidents (MVA), representing a tremendous burden to healthcare. The aim of this study was to define the prevalence of coronary disease and its possible role in motor-vehicle accidents. METHODS AND RESULTS: We examined consecutive cases of non-hospital sudden death autopsies in the area of West Quebec during the period of 2002-2006, and we focused on those victims of MVA. Severe coronary artery disease (CAD) was defined as a narrowing of ≥ 75% of a cross-sectional area or the presence of acute plaque events in major epicardial coronary arteries. From a total cohort of 1260 autopsies, MVA were responsible for 123 deaths, 100 of whom were men and 23 were women. Significant CAD was documented in approximately 37% of these cases. In individuals older than 60 years, the prevalence of significant CAD and ischemia were 86.2% and 19.8%, respectively. A percentage of 40% of the coronary patients showed erratic driving before the accident, as observed by witnesses. Statistical analysis showed that an individual affected by CAD has 9% probability of suffering a motor-vehicle accident. CONCLUSIONS: The prevalence of severe CAD and acute myocardial ischemia is very high among individuals who have suffered a MVA. Our data suggest the hypothesis that acute CAD could be the cause of accidents in a large group of the drivers affected by coronary disease. For these reasons CAD could be investigated in drivers above 50 years old, as a possible preventive measure and determinant of individual risk stratification.