Characterizing double-back stutter in low to multi-copy number regimes in forensically relevant STR loci
Sheehan, Jennifer Lee
MetadataShow full item record
Modern DNA analysis is possible due to the discovery of repeating microsatellite regions in DNA and successful implementation of the polymerase chain reaction (PCR) in laboratories. PCR amplification chemistries that contain short tandem repeat (STR) loci are sensitive. As a result, the discrimination power within human identification sciences has increased in recent years. Despite these advances, cellular admixtures are commonly collected, and the resultant “DNA mixture profile” is difficult to interpret as it is often encumbered by low-signals and allele drop-out. Regularly detected PCR artifacts can further complicate interpretation. One commonly encountered artifact is stutter, the result of strand slippage during PCR. Stutter can be of two types: forward and reverse. Reverse stutter (or back stutter) is the most prevalent and is one repeat unit shorter (n - 1) than the template strand. In contrast, forward stutter is one repeat unit longer (n + 1). If a reverse stutter amplicon is produced there is the distinct possibility that a stutter product of stutter may occur. This artifact is usually referred to as double-back stutter (DBS) or n - 2 stutter. Recently there has been renewed interest in examining signal approaching baseline levels. As the sensitivity of the process improves, so does the probability of detecting DBS. Therefore, studies that examine the peak height distributions, rarity, stutter signal-to-noise distances and the general impact of DBS on the signal are warranted. Models simulating PCR, and the entire forensic DNA process, have been created by this laboratory. The work presented herein builds upon a preexisting model; specifically, the dynamic model was extended such that DNA profiles consisting of 21 autosomal STRs, consistent with the GlobalFilerTM multiplex, are simulated. Furthermore, this expansion incorporated a three-type Galton-Watson branching process allowing DBS to be added to the simulated electropherogram (EPG). The in silico model was used to simulate the amplification of a 1:43 and 1:73 mixture at a total DNA concentration of 0.3 and 0.5 ng, respectively. We chose these extreme mixture ratios because the signal from these minor contributors would be most susceptible to DBS effects from the major contributor. A total of 1200 alleles from each contributor were simulated at each target, and effects of DBS on the signal from the minor contributor were characterized. At 0.3 and 0.5 ng both the noise and stutter signal histograms are right-skewed and a Kolmogorov-Smirnov (KS) test indicates that the noise and DBS were significantly different (p-value < 4x10-6). The average peak height of DBS for all loci in both scenarios were less than 50 RFU (Relative Fluorescence Units), and the DBS ratios ranged from 0.29 to 2.15% of the main allele, with the median ratios less than 0.5%. A per locus analytical threshold (AT) was calculated for both the 0.3 and 0.5 ng targets using two k-values: 3 and 4. The k-value is chosen based on the Type I risk assessment, wherein increasing the k-value increases AT. The percentage of DBS peaks greater than AT when k = 3 for the mixtures amplified at 0.3 and 0.5 ng ranged from 0 to 7.08% and 0 to 10.50%, respectively. Interestingly, when k = 4 the percentage of DBS peaks greater than AT for 0.3 and 0.5 ng reduced to 0 to 1.08% and 0 to 0.17%, respectively. This suggests that modeling DBS in continuous systems may not be necessary if the laboratory continues to rely on a system that requires an AT of sufficient strength. However, with the advent of Bayesian or machine learning-based approaches to analyzing EPGs, thus removing AT in its entirety, a complete understanding of the prevalence of DBS is necessary. This work shows that DBS from an extreme major using our laboratory protocols is not likely to be in the same signal regime as the signal from alleles; however, it does show that signal from DBS is significantly different from noise. Therefore, the software expert pair should be carefully considered during the validation stage and laboratories should consider DBS during interpretation, especially if enhanced post-PCR parameters are implemented into the forensic laboratory process.