Guro Dørum

Exact computation of the distribution of likelihood ratios with forensic applications.

If complex DNA profiles, conditioned on multiple individuals are evaluated, it may be difficult to assess the strength of the evidence based on the likelihood ratio. A likelihood ratio does not give information about the relative weights that are provided by separate contributors. Alternatively, the observed likelihood ratio can be evaluated with respect to the distribution of the likelihood ratio under the defense hypothesis. We present an efficient algorithm to compute an exact distribution of likelihood ratios that can be applied to any LR-based model. The distribution may have several applications, but is used here to compute a p-value that corresponds to the observed likelihood ratio. The p-value is the probability that a profile under the defense hypothesis, substituted for a questioned contributor e.g. suspect, would attain a likelihood ratio which is at least the same magnitude as that observed. The p-value can be thought of as a scaled version of the likelihood ratio, giving a quantitative measure of the strength of the evidence relative to the specified hypotheses and the model used for the analysis. The algorithm is demonstrated on examples based on real data. R code for the algorithm is freely available in the R package euroMix.

In vitro comparison of commensal, probiotic and pathogenic strains of Enterococcus faecalis

In vivo studies have provided evidence that micro-organisms have important roles in immunological, digestive and respiratory functions, conferring health benefits on the host. Several in vitro methods have been advised for the initial screening of microbes with potential health effects. The objective of the present study was to employ such in vitro methodology to characterise different strains of Enterococcus faecalis. The characteristics of a commercial product marketed as a probiotic, Symbioflor-1 (Symbiopharm), were compared with the characteristics of both pathogenic and commensal strains. Tolerance towards low pH and viability after exposure to human gastric and duodenal juices were assayed. Symbioflor-1 was the most susceptible strain to these treatments when compared with the other E. faecalis strains. Furthermore, Symbioflor-1 exhibited the lowest adhesion capacity to intestinal epithelial cells (IEC) and mucus. Competitive binding studies using heparin indicated that glycosaminoglycans might be involved in the adhesion to IEC, but also that differences in these putative bacteria-host interactions do not cause the relative low adhesion capacity of Symbioflor-1. Maturation of dendritic cells (DC) after exposure to bacteria was assayed as an indication of an immunomodulatory effect. All strains induced a moderate elevation of the DC maturation markers CD83 and CD86; however, no strain-specific differences were detected. Correlations between in vitro and in vivo studies are discussed. Although in vitro assaying is a rational starting point for the selection of microbes with a potential health benefit, it is emphasised that human clinical trials are the definite tool for establishing probiotic status.

Database extraction strategies for low-template evidence.

Often in forensic cases, the profile of at least one of the contributors to a DNA evidence sample is unknown and a database search is needed to discover possible perpetrators. In this article we consider two types of search strategies to extract suspects from a database using methods based on probability arguments. The performance of the proposed match scores is demonstrated by carrying out a study of each match score relative to the level of allele drop-out in the crime sample, simulating low-template DNA. The efficiency was measured by random man simulation and we compared the performance using the SGM Plus kit and the ESX 17 kit for the Norwegian population, demonstrating that the latter has greatly enhanced power to discover perpetrators of crime in large national DNA databases. The code for the database extraction strategies will be prepared for release in the R-package forensim.

Models and implementation for relationship problems with dropout

Allelic dropout in relationship problems may commonly appear in areas such as disaster victim identification and the identification of missing persons. If dropout is not accounted for, the results may be incorrect interpretation of profiles, loss of valuable information and biased results. In this paper, we explore different models for dropout in kinship cases and present an efficient implementation for one of the models. The implementation allows for dropout to be handled simultaneously with phenomena like silent alleles and mutations that may also cause discordances in relationship data, in addition to subpopulation correction. The implemented dropout model is freely available in the new version of the Familias software. The concepts and methods are illustrated on real and simulated data.

Mixtures with relatives: A pedigree perspective

DNA mixture evidence pertains to cases where several individuals may have contributed to a biological stain. Statistical methods and software for such problems are available and a large number of cases can be handled adequately. However, one class of mixture problems remains untreated in full generality in the literature, namely when the contributors may be related. Disregarding a plausible close relative of the perpetrator as an alternative contributor (identical twin is the most extreme case) may lead to overestimating the evidence against a suspect. Existing methods only accommodate pairwise relationships such as the case where the suspect and the victim are siblings, for example. In this paper we consider relationships in full generality, conveniently represented by pedigrees. In particular, these pedigrees may involve inbreeding, for instance when the parents of an individual of interest are first cousins. Furthermore our framework handles situations where the opposing parties in a court case (prosecution and defence) propose different family relationships. Consequently, our approach combines classical mixture and kinship problems. The basic idea of this paper is to formulate the problem in a way that allows for the exploitation of currently available methods and software designed originally for linkage applications. We have developed a freely available R package, euroMix based on another package, paramlink, and we illustrate the ideas and methods on real and simulated data.

Relationship inference based on DNA mixtures

Today, there exists a number of tools for solving kinship cases. But what happens when information comes from a mixture? DNA mixtures are in general rarely seen in kinship cases, but in a case presented to the Norwegian Institute of Public Health, sample DNA was obtained after a rape case that resulted in an unwanted pregnancy and abortion. The only available DNA from the fetus came in form of a mixture with the mother, and it was of interest to find the father of the fetus. The mother (the victim), however, refused to give her reference data and so commonly used methods for paternity testing were no longer applicable. As this case illustrates, kinship cases involving mixtures and missing reference profiles do occur and make the use of existing methods rather inconvenient. We here present statistical methods that may handle general relationship inference based on DNA mixtures. The basic idea is that likelihood calculations for mixtures can be decomposed into a series of kinship problems. This formulation of the problem facilitates the use of kinship software. We present the freely available R package relMix which extends on the R version of Familias. Complicating factors like mutations, silent alleles, and θ-correction are then easily handled for quite general family relationships, and are included in the statistical methods we develop in this paper. The methods and their implementations are exemplified on the data from the rape case.

Mixtures with relatives and linked markers

Mixture DNA profiles commonly appear in forensic genetics, and a large number of statistical methods and software are available for such cases. However, most of the literature concerns mixtures where the contributors are assumed unrelated and the genetic markers are unlinked. In this paper, we consider mixtures of linked markers and related contributors. If no relationships are involved, linkage can be ignored. While unlinked markers can be treated independently, linkage introduces dependencies. The use of linked markers presents statistical and computational challenges, but may also lead to a considerable increase in power since the number of markers available is much larger if we do not require the markers to be unlinked. In addition, some cases that cannot be solved with an unlimited number of unlinked autosomal markers can be solved with linked markers. We focus on two special cases of linked markers: pairs of linked autosomal markers and X-chromosomal markers. A framework is presented for calculation of likelihood ratios for mixtures with general relationships and with linkage between any number of markers. Finally, we explore the effect of linkage disequilibrium, also called allelic association, on the likelihood ratio.

Genome wide transcriptional profiling of acclimation to photoperiod in high-latitude accessions of Arabidopsis thaliana

Three Arabidopsis thaliana accessions originating from the northernmost boundary of the species distribution in Norway (59-68°N) were used to study global wide transcriptional responses to 16 and 24 h photoperiods during flower initiation. Significant analysis of microarrays (SAM), analyses of statistically overrepresented gene ontologies (GOstat) and gene set enrichment analyses (GSEA) were used to identify candidate genes and genetic pathways underlying phenotypic adaptations of accessions to different photoperiods. Statistical analyses identified 732 and 258 differentially expressed genes between accessions in 16 and 24 h photoperiod, respectively. Among significantly expressed genes, ethylene mediated signaling pathway was significantly overrepresented in 16 h photoperiod, while genes involved in response to auxin stimulus were found to be significantly overrepresented in 24 h photoperiod. Several gene sets were found to be differentially expressed among accessions, e.g. cold acclimation, dehydration response, phytochrome signaling, vernalization response and circadian clock regulated flowering time genes. These results revealed several candidate genes and pathways likely involved in transcriptional control of photoperiodic response. In particular, ethylene and auxin signaling pathway may represent candidate genes contributing to local adaptation of high-latitude accessions of A. thaliana.

High-throughput genotyping of unknown genomic terrain in complex plant genomes: lessons from a case study.

Novel high-throughput genotyping technologies have facilitated rapid genotyping of single nucleotide polymorphisms in non-model organisms. Most plant species have complex genomes with a large proportion of their genes having one or more paralogous copies due to single gene duplications and ancient or recent polyploidization events. These paralogous gene copies are potential sources of genotyping errors, and hence genotyping of plant genomes is inherently difficult. Here we present a case study that exemplifies paralog-related problems in high-throughput genotyping of plant genomes. We used the MassARRAY genotyping platform to genotype the LpIRI locus in L. perenne populations; this gene is thought to be involved in low-temperature stress tolerance. The dissection of the molecular genetics underlying the genotyping results provides a good example of how unknown paralogs can mask the true genotype of the locus, instructive to the non-specialist plant researcher and breeder.

Rotation gene set testing for longitudinal expression data

Gene set analysis methods are popular tools for identifying differentially expressed gene sets in microarray data. Most existing methods use a permutation test to assess significance for each gene set. The permutation tests assumption of exchangeable samples is often not satisfied for time-series data and complex experimental designs, and in addition it requires a certain number of samples to compute p-values accurately. The method presented here uses a rotation test rather than a permutation test to assess significance. The rotation test can compute accurate p-values also for very small sample sizes. The method can handle complex designs and is particularly suited for longitudinal microarray data where the samples may have complex correlation structures. Dependencies between genes, modeled with the use of gene networks, are incorporated in the estimation of correlations between samples. In addition, the method can test for both gene sets that are differentially expressed and gene sets that show strong time trends. We show on simulated longitudinal data that the ability to identify important gene sets may be improved by taking the correlation structure between samples into account. Applied to real data, the method identifies both gene sets with constant expression and gene sets with strong time trends.