A comparison of likelihood ratios calculated from surface DNA mixtures using MPS and CE Technologies.

This study evaluates the performance of analysing surface DNA samples using massively parallel sequencing (MPS) compared to traditional capillary electrophoresis (CE). A total of 30 samples were collected from various surfaces in an office environment and were analysed with CE and MPS. These were compared against 60 reference samples (office inhabitants). To identify contributors, likelihood ratios (LRs) were calculated for MPS and CE data using the probabilistic genotyping software MPSproto and EuroForMix respectively. Although a higher number of sequences/peaks were observed per DNA profile in MPS compared to CE, LR values were found to be lower for MPS data formats. This might be the result of the increased complexity of MPS data, along with a possible elevation of unknown alleles and/or artefacts. The study highlights avenues for improving MPS data quality and analysis to facilitate more robust interpretation of challenging casework-like samples.

An overview of autosomal STRs and identity SNPs in a Norwegian population using massively parallel sequencing.

In recent years, probabilistic genotyping software has been adapted for the analysis of massively parallel sequencing (MPS) forensic data. Likelihood ratios (LR) are based on allele frequencies selected from populations of interest. This study provides an outline of sequence-based (SB) allele frequencies for autosomal short tandem repeats (aSTRs) and identity single nucleotide polymorphisms (iSNPs) in 371 individuals from Southern Norway. 27 aSTRs and 94 iSNPs were previously analysed with the ForenSeq™ DNA Signature Prep Kit (Verogen). The number of alleles with frequencies less than 0.05 for sequenced-based alleles was 4.6 times higher than for length-based alleles. Consistent with previous studies, it was observed that sequence-based data (both with and without flanks) exhibited higher allele diversity compared to length-based (LB) data; random match probabilities were lower for SB alleles confirming their advantage to discriminate between individuals. Two alleles in markers D22S1045 and Penta D were observed with SNPs in the 3´ flanking region, which have not been reported before. Also, a novel SNP with a minor allele frequency (MAF) of 0.001, was found in marker TH01. The impact of the sample size on minor allele frequency (MAF) values was studied in 88 iSNPs from Southern Norway (n = 371). The findings were then compared to a larger Norwegian population dataset (n = 15,769). The results showed that the smaller Southern Norway dataset provided similar results, and it was a representative sample. Population structure was analyzed for regions within Southern Norway; FST estimates for aSTR and iSNPs did not indicate any genetic structure. Finally, we investigated the genetic differences between Southern Norway and two other populations: Northern Norway and Denmark. Allele frequencies between these populations were compared, and we found no significant frequency differences (p-values > 0.0001). We also calculated the pairwise FST values per marker and comparisons between Southern and Northern Norway showed small differences. In contrast, the comparisons between Southern Norway and Denmark showed higher FST values for some markers, possibly driven by distinct alleles that were present in only one of the populations. In summary, we propose that allele frequencies from each population considered in this study could be used interchangeably to calculate genotype probabilities.

MPSproto: An extension of EuroForMix to evaluate MPS-STR mixtures.

We have developed MPSproto as an extension of EuroForMix to improve handling of stutter artefacts and other typing errors that commonly occur in MPS-STR data. MPSproto implements two models for read depth: gamma and negative binomial. It differs from EuroForMix in that calibration is required before mixtures are interpreted. In this study a mixture dataset (2-4 persons) was revisited, where EuroForMix interpretation of MPS-STR mixtures using the LUS+ format was first described; the performance of this model was compared to the MPSproto models. Results indicated that, overall, the MPSproto models performed better than the conventional EuroForMix model, and the gamma model implemented in MPSproto performed best. Differences were highlighted and further investigated to establish causality. Goodness of fit tests showed that the MPSproto models were adequate for the sequence reads when a low analytical threshold was applied.

A comprehensive characterization of MPS-STR stutter artefacts.

Interpretation of DNA evidence involving mixtures is challenging when alleles from minor contributors coincide with stutters from major contributors. To accommodate this, it is important to have a good understanding of stutter sequence formation trends. Here, multiple stutter types were characterized based on massively parallel sequencing (MPS) data from 387 single source samples, using the Verogen ForenSeq™ DNA Signature Prep kit. A beta regression model was used to investigate the relationship between the stutter proportion and candidate explanatory variables. In the final model, stutter proportions were explained by the length of the parental uninterrupted stretch (PTUS), which is comparable to block length of the missing motif (BLMM). Also, different stutter types (n+1, n-1, n+2, n-2, n0) were analyzed separately per locus. The fitted stutter models were then integrated into an extended probabilistic genotyping model based on EuroForMix (MPSproto). An illustrative minor/major mock mixture example is discussed. Evaluation of multiple types of stutters on a per locus basis improved the probabilistic genotyping result compared to the conventional EuroForMix model, using the LUS+ nomenclature.

Parallel CRISPR-Cas9 screens clarify impacts of p53 on screen performance.

CRISPR-Cas9 genome engineering has revolutionised high-throughput functional genomic screens. However, recent work has raised concerns regarding the performance of CRISPR-Cas9 screens using TP53 wild-type human cells due to a p53-mediated DNA damage response (DDR) limiting the efficiency of generating viable edited cells. To directly assess the impact of cellular p53 status on CRISPR-Cas9 screen performance, we carried out parallel CRISPR-Cas9 screens in wild-type and TP53 knockout human retinal pigment epithelial cells using a focused dual guide RNA library targeting 852 DDR-associated genes. Our work demonstrates that although functional p53 status negatively affects identification of significantly depleted genes, optimal screen design can nevertheless enable robust screen performance. Through analysis of our own and published screen data, we highlight key factors for successful screens in both wild-type and p53-deficient cells.

The invention of CRISPR-Cas9 genome editing has unlocked a greater understanding of the human genome. Researchers can use this system to make targeted cuts in any gene in the genome, forcing the cell to perform a rapid repair at the cut site. These repairs often introduce mutations into the damaged area, adding or removing DNA letters and disrupting the gene. This allows researchers to study what happens to cells when specific genes are missing, which can help to uncover what each gene is for. One of the most comprehensive ways to use this technique is to perform a CRISPR-Cas9 screen, which disrupts each gene in the genome one by one. For a CRISPR-Cas9 screen to work well, a cell needs to survive the cuts to its genome. But there is a crucial gene that can stop this happening. Often described as the 'guardian of the genome', this gene codes for a protein called p53, a tumour suppressor that helps to stop a cell turning cancerous when its DNA becomes damaged. This protein activates when the cell senses a cut in its genetic material and can kill the cell if it fails to make a successful repair. Recent work has shown that the presence of a working copy of the gene for the p53 protein might limit the ability of CRISPR-Cas9 to edit genes. But the evidence was inconclusive. So, Bowden, Morales-Juarez et al. performed two parallel CRISPR-Cas9 screens in human cells with and without p53 to find out more. This revealed that CRISPR-Cas9 can inactivate genes in both normal cells and cells lacking the p53 protein, but that it works better in cells without p53. This was because, when p53 was active, the cells initiated a protective response against the CRISPR-Cas9 cuts. This changed the patterns of genes successfully inactivated by the screen, but it did not make the results unusable. Careful experimental design and thorough data analysis made it possible to get useful results even in cells with functional p53 protein. The gene for p53 has mutations in around half of human cancers. So, understanding how it affects CRISPR-Cas9 screens could influence the design of future experiments. It is possible that the effects of the p53 protein could vary from cell type to cell type, and with different p53 mutations. Comparisons like the one performed here could help to further unpick how the cell's DNA repair systems might interfere with future CRISPR experiments.