Biology subjects

Eriksen, P. S.

Publications and source records attributed to Eriksen, P. S..

5 recordsLinked to original sources

bioRxiv2025

Likelihood ratio estimation of partial Y-STR profile matches using discrete Laplace models and marginalisation

The discrete Laplace model can be used to estimate the evidential weight of a match between the Y-chromosome short tandem repeat (Y-STR) DNA profiles of the evidence material and a suspect. When the weight of evidence of a match between partial Y-STR profiles is assessed, a discrete Laplace model, restricted to the observed loci in the partial evidence profile, is constructed. However, fitting such a discrete Laplace model is time-consuming since it requires estimating the parameters of the discrete Laplace model and validating the model. We implemented the marginalisation method for discrete Laplace models to estimate the evidential weight of a match between a partial Y-STR evidence profile and a Y-STR reference profile. Since the marginalisation method is based on discrete Laplace models for complete Y-STR profiles, no additional models need to be constructed. We compared the likelihood ratios obtained with (1) the marginalisation method and (2) by fitting new discrete Laplace models restricted to the observed loci in the partial evidence profile. In most cases, the two methods yielded similar likelihood ratios.

genetics↗

bioRxiv2025

Probabilities of finding trace profile donors and their paternal relatives in Y-STR reference databases

Forensic investigative genetic genealogy using Y-chromosome short tandem repeat (Y-STR) DNA profiles can give investigative leads in criminal cases by searching for the Y-STR trace profile or similar but not identical Y-STR profiles in relevant Y-STR databases. We conducted a simulation study with Yfiler Plus and PowerPlex(R) Y23 Y-STR profiles to estimate the probabilities of finding matches and near-matches in Y-STR databases. The success rate of finding the trace profile donors or their close relatives was quantified. We used the malan R software package to simulate the populations based on the Wright-Fisher model with the YHRD Y-STR mutation rates where uncertainties were incorporated in a Bayesian manner, a variance in reproductive success of 0.2, and a constant size for 100 generations followed by a 2% growth for 150 generations. Y-STR databases were generated by randomly drawing Y-STR profiles from a Yfiler Plus and PowerPlex(R) Y23 Y-STR population data set, respectively. In a population of 500,066 individuals, a database size of 0.5% of the population resulted in a Y-STR database match probability of ca. 6% and 10% for Yfiler Plus and PowerPlex(R) Y23, respectively. Increasing the database size to 5% of the population resulted in a Y-STR match probability of ca. 41% and 54% for Yfiler Plus and PowerPlex(R) Y23, respectively. When a Y-STR match was found in the database, the probability of one of the individuals with the matching profiles being related within five meioses to the trace donor was ca. 64% and 56% for Yfiler Plus and PowerPlex(R) Y23, respectively, including the cases where the Y-STR profile originated from the donor. In this case, the closest relative in the database was found among the matching individuals with a probability of ca. 91%.

genetics↗

bioRxiv2024

Enhanced SNP Genotyping with Symmetric Multinomial Logistic Regression

In genotyping, determining Single Nucleotide Polymorphisms (SNPs) is standard practice, but it becomes difficult when analysing small quantities of input DNA, as is often required in forensic applications. Existing SNP genotyping methods, such as the HID SNP Genotyper Plugin (HSG) from Thermo Fisher Scientific, perform well with adequate DNA input levels but often produce erroneously called genotypes when DNA quantities are low. To mitigate these errors, genotype quality can be checked with the HSG. However, enforcing the HSGs quality checks decreases the call rate by introducing more no-calls, and it does not eliminate all wrong calls. This study presents and validates a Symmetric Multinomial Logistic Regression (SMLR) model designed to enhance genotyping accuracy and call rate with small amounts of DNA. Comprehensive bootstrap and cross-validation analyses across a wide range of DNA quantities demonstrate the robustness and efficiency of the SMLR model in maintaining high call rates without compromising accuracy compared to the HSG. For DNA amounts as low as 31.25 pg, the SMLR method reduced the rate of no-calls by 50.0% relative to the HSG while maintaining the same rate of wrong calls, resulting in a call rate of 96.0%. Similarly, SMLR reduced the rate of wrong calls by 55.6% while maintaining the same call rate, achieving an accuracy of 99.775%. The no-call and wrong-call rates were significantly reduced at 62.5-250 pg DNA. The results highlight the SMLR models utility in optimising SNP genotyping at suboptimal DNA concentrations, making it a valuable tool for forensic applications where sample quantity and quality may be decreased. This work reinforces the feasibility of statistical approaches in forensic genotyping and provides a framework for implementing the SMLR method in practical forensic settings. The SMLR model applies for genotyping biallelic data with a signal (e.g. reads, counts, or intensity) for each allele. The model can also improve the allele balance quality check.

genetics↗

bioRxiv2022

Weight of evidence of Y-STR matches computed with the discrete Laplace method: Impact of adding a suspect's profile to a reference database

The discrete Laplace method is recommended by multiple parties (including the International Society of Forensic Genetics, ISFG) to estimate the weight of evidence in criminal cases when a suspects Y-STR profile matches the crime scene Y-STR profile. Unfortunately, modelling the distribution Y-STR profiles in the database is time-consuming and requires expert knowledge. When the suspects Y-STR profile is added to the database, as would be the protocol in many cases, the discrete Laplace model must be recomputed. We found that the likelihood ratios with and without adding the suspects Y-STR profile were almost identical with 1,000 or more Y-STR profiles in the database for Y-STR profiles with 8, 12, and 17 loci. Thus, likelihood ratio calculations can be performed in seconds if a an established discrete Laplace model based on at least 1,000 Y-STR profiles is used. A match in a database with 17 Y-STR loci from at least 1,000 male individuals results in a likelihood ratio above 10,000 in approximately 94% of the cases, and above 100,000 in approximately 82% of the cases. We offer a freely available IT tool for estimating the discrete Laplace model of the STR profiles in a database and the likelihood ratio. HighlightsO_LIThe discrete Laplace method is suitable for estimating the weight of evidence of matches with 17 Y-STRs. C_LIO_LILRs based on the discrete Laplace method are 10-100 times higher (in median) than those based on Brenners{kappa} method. C_LIO_LIA database with 17 STRs from at least 1,000 males gives LRs of above 10,000 in approximately 94% of the cases and above 100,000 in approximately 82% of the cases with the discrete Laplace method. C_LIO_LIThe weight of evidence of a matching Y-STR profile is computed within seconds and easily documented when a precomputed discrete Laplace model is available (an IT tool is provided). C_LIO_LI50% of all Yfiler Plus matches are between male relatives within a genetic distance of five meioses. C_LI

genetics↗

bioRxiv2022

SNP calling for the Illumina Infinium Omni5-4 SNP BeadChip kit using the butterfly method

We introduce the "butterfly method" for SNP calling with the Illumina Infinium Omni5-4 BeadChip kit without the use of Illumina GenomeStudio software. The method is a within-sample method and does not use other samples nor population frequencies to call SNPs. The butterfly method is based on a three-component mixture of normal distributions, in which parameters are easily found using the open-source statistical software R. This makes the method transparent, straight-forward to change parameters according to the users needs, and easy to analyse the data within R after the SNPs have been called. We contribute with two open-source R packages that make SNP calling easy by helping with bookkeeping and by giving easy access to meta-information about the SNPs on the Illumina Infinium Omni5-4 BeadChip Kit (including chromosome, probe type, and SNP bases). We test our method on > 4 mio. SNPs and compare the results with those obtained with the GenTrain method used by Illumina GenomeStudio as well as SNPs obtained by PCR-free whole genome sequencing (WGS). We demonstrate two variants of our method: one where we account for potential probe type bias by estimating a separate model for each probe type (type I and type II) and another that uses a general model such that the models parameter estimates do not depend on the sample that is being analysed. We focused on varying the no-call rate and show how it changed the concordance with that of WGS. This is done by using a threshold on the a posteriori probability of belonging to a SNP cluster and by using the number of beads to adjust the stringency of the no-call mechanism. With the butterfly method, we achieve a SNP call rate of around 99% and a SNP concordance of around 99% with the WGS data. By lowering the a posteriori probability threshold for no-calls, we can get a higher call rate fraction than the GenomeStudio and by using a higher a posteriori probability threshold, we can achieve a higher concordance with the WGS data than the GenomeStudio.

bioinformatics↗