Robust detection of individual forensic profiles in DNA mixtures

Targeted sequencing of high-density SNPs provides an enhanced tool for forensic applications and genetic landscape exploration in Chinese Korean ethnic group

BACKGROUND

In this study, we present a NGS-based panel designed for sequencing 1993 SNP loci for forensic DNA investigation. This panel addresses unique challenges encountered in forensic practice and allows for a comprehensive population genetic study of the Chinese Korean ethnic group. To achieve this, we combine our results with datasets from the 1000 Genomes Project and the Human Genome Diversity Panel.

RESULTS

We demonstrate that this panel is a reliable tool for individual identification and parentage testing, even when dealing with degraded DNA samples featuring exceedingly low SNP detection rates. The performance of this panel for complex kinship determinations, such as half-sibling and grandparent-grandchild scenarios, is also validated by various kinship simulations. Population genetic studies indicate that this panel can uncover population substructures on both global and regional scales. Notably, the Han population can be distinguished from the ethnic minorities in the northern and southern regions of East Asia, suggesting its potential for regional ancestry inference. Furthermore, we highlight that the Chinese Korean ethnic group, along with various Han populations from different regional areas and certain northern ethnic minorities (Daur, Tujia, Japanese, Mongolian, Xibo), exhibit a higher degree of genetic affinities when examined from a genomic perspective.

CONCLUSION

This study provides convincing evidence that the NGS-based panel can serve as a reliable tool for various forensic applications. Moreover, it has helped to enhance our knowledge about the genetic landscape of the Chinese Korean ethnic group.

Application of SNPs with low minor allele frequencies in missing person identification (MPI) through kinship analysis of DNA mixtures

The need to identify a missing person (MP) through kinship analysis of DNA samples found at a crime scene has become increasingly prevalent. DNA samples from MPs can be severely degraded, contain little DNA and mixed with other contributors, which often makes it difficult to apply conventional methods in practice. This study developed a massively parallel sequencing-based panel that contains 1661 single-nucleotide polymorphisms (SNPs) with low minor allele frequencies (MAFs) (averaged at 0.0613) in the Chinese Han population, and the strategy for relationship inference from DNA mixtures comprising different numbers of contributors (NOCs) and of varying allele dropout probabilities. Based on the simulated dataset and genotyping results of 42 artificial DNA mixtures (NOC = 2-4), it was observed that the present SNP panel was sufficient for balanced mixtures when referenced to the closest relatives (parents/offspring and full siblings). When the mixture profiles suffered from dropout, incorrect assignments were markedly associated with relatedness, NOC and the dropout level. We, therefore, indicate that SNPs with low MAFs could be reliably interpreted for MP identification through the kinship analysis of complex DNA mixtures. Further studies should be extended to more possible scenarios to test the feasibility of this present approach.

Evaluation of large-scale highly polymorphic microhaplotypes in complex DNA mixtures analysis using RMNE method

DNA mixture interpretation is one of the most challenging problems in forensics. Complex DNA mixtures are more difficult to analyze when there are more than two contributors or related contributors. Microhaplotypes (MHs) are polymorphic genetic markers recently discovered and employed in DNA mixture analysis. However, the evidentiary interpretation of the MH genotyping data needs more debate. The Random Man Not Excluded (RMNE) method analyzes DNA mixtures without using allelic peak height data or the number of contributors (NoC) assumptions. This study aimed to assess how well RMNE interpreted mixed MH genotyping data. We classified the MH loci from the 1000 Genomes Project database into groups based on their Ae values. Then we performed simulations of DNA mixtures with 2-10 unrelated contributors and DNA mixtures with a pair of sibling contributors. For each simulated DNA mixture, incorrectly included ratios were estimated for three types of non-contributors: random men, parents of contributors, and siblings of contributors. Meanwhile, RMNE probability was calculated for contributors and three types of non-contributors, allowing loci mismatch. The results showed that the MH number, the MH Ae values, and the NoC affected the RMNE probability of the mixture and the incorrectly included ratio of non-contributors. When there were more MHs, MHs with higher Ae values, and a mixture with less NoC, the RMNE probability, and the incorrectly included ratio decreased. The existence of kinship in mixtures complicated the mixture interpretation. Contributors' relatives as non-contributors and related contributors in the mixture increased the demands on the genetic markers to identify the contributors correctly. When 500 highly polymorphic MHs with Ae values higher than 5 were used, the four individual types could be distinguished according to the RMNE probabilities. This study reveals the promising potential of MH as a genetic marker for mixed DNA interpretation and the broadening of RMNE as a parameter indicating the relationship of a specific individual with a DNA mixture in the DNA database search.

A novel computational strategy to predict the value of the evidence in the SNP-based forensic mixtures

This study introduces a methodology for inferring the weight of the evidence (WoE) in the single nucleotide polymorphism (SNP)-typed DNA mixtures of forensic interest. First, we redefined some algebraic formulae to approach the semi-continuous calculation of likelihoods and likelihood ratios (LRs). To address the allelic dropouts, a peak height ratio index (“h,” an index of heterozygous state plausibility) was incorporated into semi-continuous formulae to act as a proxy for the “split-drop” model of calculation. Second, the original ratio at which a person of interest (POI) has entered into the mixture was inferred by evaluating the DNA amounts conferred by unique genotypes to any possible permutation of any locus of the typing protocol (unique genotypes are genotypes that appear just once in the relevant permutation). We compared this expected ratio (MR_ex) to all the mixing ratios emerging at all other permutations of the mixture (MR_obs) using several (1 - χ²) tests to evaluate the probability of each permutation to exist in the mixture according to quantitative criteria. At the level of each permutation state, we multiplied the (1 - χ²) value to the genotype frequencies and the h index. All the products of all the permutation states were finally summed to give a likelihood value that accounts for three independent properties of the mixtures. Owing to the (1 - χ²) index and the h index, this approach qualifies as a fully continuous methodology of LR calculation. We compared the MRs and LRs emerging from our methodology to those generated by the EuroForMix software ver. 3.0.3. When the true contributors were tested as POIs, our procedure generated highly discriminant LRs that, unlike EuroForMix, never overcame the corresponding single-source LRs. When false contributors were tested as POIs, we obtained a much lower LR value than that from EuroForMix. These two findings indicate that our computational method is more reliable and realistic than EuroForMix.

Analysis of complex DNA mixtures using massively parallel sequencing of SNPs with low minor allele frequencies

DNA mixtures from 3 or more contributors have proven difficult to analyze using the current state-of-the-art method of short-tandem repeat (STR) amplification followed by capillary electrophoresis (CE). Here we analyze samples from both laboratory-defined mixtures and complex multi-contributor touch samples using a single nucleotide polymorphism (SNP) panel comprised of 2311 low-minor-allele-frequency loci, combined with massively parallel sequencing (MPS). This approach demonstrates that as many as 10 people can be identified in touch samples using a threshold of -Log P(RMNE) of 6, and a detection rate of 18-94 % across 10 different materials using a threshold of -Log P(RMNE) of 2. Thirty-two false positives were observed in 100 touch samples.

The advances in DNA mixture interpretation

In forensic genetics, the analysis of DNA in biological samples is a valuable tool for personal identification. There is an increasing demand in analyzing of the mixed DNA which may provide insightful investigative instructions. With the continuous effort for the improvement of individual identification, complicated mixed stains represent a growing fraction of the samples processed by forensic laboratories. Recent technological advances have enabled quantitative analysis of DNA mixture and emerging sequencing approaches to decipher the complicated DNA mixture. Here, we describe the use of different genetic markers, typing approaches and analytical methods in mixture analysis, and how useful information can be obtained from complicated DNA mixture.

Separation/extraction, detection, and interpretation of DNA mixtures in forensic science (review)

Interpreting mixed DNA samples containing material from multiple contributors has long been considered a major challenge in forensic casework, especially when encountering low-template DNA (LT-DNA) or high-order mixtures that may involve missing alleles (dropout) and unrelated alleles (drop-in), among others. In the last decades, extraordinary progress has been made in the analysis of mixed DNA samples, which has led to increasing attention to this research field. The advent of new methods for the separation and extraction of DNA from mixtures, novel or jointly applied genetic markers for detection and reliable interpretation approaches for estimating the weight of evidence, as well as the powerful massively parallel sequencing (MPS) technology, has greatly extended the range of mixed samples that can be correctly analyzed. Here, we summarized the investigative approaches and progress in the field of forensic DNA mixture analysis, hoping to provide some assistance to forensic practitioners and to promote further development involving this issue.

Evaluation of the Microhaplotypes panel for DNA mixture analyses

The identification of a suspect in a DNA mixture typed with the standard short tandem repeat polymorphism (STR) kits has faced challenges. Several improved methods or technologies have been introduced to address this issue. However, some complex situations in the process remain elusive. In the present study, we presented a panel of 26 tiny microhaplotypes, each generating a relatively high (>3.0) effective number of alleles (A_e) and possessing low (<50 bp) sequence lengths. The average A_e and heterozygosity values among the 9 populations of 26 microhaps were in ranges from 2.60 to 4.54 and 0.59 to 0.96, respectively. Power of discrimination and power of exclusion values were ranged from 0.49 to 0.87 and 0.29 to 0.94, respectively. Significant positive correlations have been found between A_e values and heterozygosity (r = 0.43, p = 0.02) or power of discrimination values (r = 0.55, p = 0.003), respectively. The cumulative probability of detecting a mixture of two unrelated individuals could reach 0.9999998 when using a panel of 26 microhaps with A_e = 3. We further tested the panel by using massively parallel sequencing, and 14 out of 26 microhaps were successfully genotyped in a single multiplex system. 60 unrelated Chinese Han individuals and 2 artificially prepared samples mixed by two unrelated contributors (in duplicate, ie. 4 mixtures) were sequenced. Approximately 32.14% of the 14 loci presented three or four alleles in the two mixtures. The likelihood ratio values to cognizance the mixtures' contributor were in a range from 1.95 × 10⁶ to 1.10 × 10⁷. The results demonstrated that the present panel could offer a valuable complementary tool in forensic applications.

Significance evaluation in factor graphs

BACKGROUND

Factor graphs provide a flexible and general framework for specifying probability distributions. They can capture a range of popular and recent models for analysis of both genomics data as well as data from other scientific fields. Owing to the ever larger data sets encountered in genomics and the multiple-testing issues accompanying them, accurate significance evaluation is of great importance. We here address the problem of evaluating statistical significance of observations from factor graph models.

RESULTS

Two novel numerical approximations for evaluation of statistical significance are presented. First a method using importance sampling. Second a saddlepoint approximation based method. We develop algorithms to efficiently compute the approximations and compare them to naive sampling and the normal approximation. The individual merits of the methods are analysed both from a theoretical viewpoint and with simulations. A guideline for choosing between the normal approximation, saddle-point approximation and importance sampling is also provided. Finally, the applicability of the methods is demonstrated with examples from cancer genomics, motif-analysis and phylogenetics.

CONCLUSIONS

The applicability of saddlepoint approximation and importance sampling is demonstrated on known models in the factor graph framework. Using the two methods we can substantially improve computational cost without compromising accuracy. This contribution allows analyses of large datasets in the general factor graph framework.